<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/">
<channel>
<title>Magpie AI articles</title>
<link>https://magpieaudit.com/articles</link>
<atom:link href="https://magpieaudit.com/articles/feed.xml" rel="self" type="application/rss+xml"/>
<description>Plain, practical articles on landing pages that convert: first screens, calls to action, pricing, proof, and being found in Google and AI answers</description>
<language>en-gb</language>
<lastBuildDate>Tue, 17 Nov 2026 23:00:00 GMT</lastBuildDate>
<item><title>What your site shows when someone shares its link</title><link>https://magpieaudit.com/articles/shared-link-previews</link><guid isPermaLink="true">https://magpieaudit.com/articles/shared-link-previews</guid><pubDate>Tue, 17 Nov 2026 23:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Search and AI</category><description>A shared link&apos;s card is drawn from your page&apos;s Open Graph tags (og:title, og:description, og:image and og:url, plus twitter:card for X). Each app fetches your page itself without running its scripts, so the tags must be in the HTML your server sends. One 1200 by 630 image under 600KB suits every big platform, and a new image address gets past an old cached picture.</description><content:encoded><![CDATA[<p>Paste a link into LinkedIn, Slack or a WhatsApp chat and a second later a card appears under it, with a picture, a title and a line or two of grey text. Your site didn&#39;t send that card. The app went off and fetched your page, read a handful of tags in its head and drew the card from those, and when it finds nothing it can use, you get a bare blue link or an empty grey box.</p>
<p>Often it isn&#39;t you sharing it, either. It&#39;s a colleague dropping you into a Slack channel, or someone chirping about you in a group chat, and the card is all anyone sees of you before they decide whether to tap.</p>
<h2>The tags a link preview reads</h2>
<p>Every platform here reads the same few Open Graph tags in your page&#39;s <code>&lt;head&gt;</code>. The <a href="https://ogp.me/">Open Graph protocol</a> asks for four on every page (<code>og:title</code>, <code>og:type</code>, <code>og:image</code> and <code>og:url</code>) and offers <code>og:description</code> for &quot;a one to two sentence description&quot;. LinkedIn&#39;s help page names title, image, description and url as the ones it needs. Here&#39;s the set I&#39;d put on brewlog.example, a made-up stock app for small coffee roasters:</p>
<pre><code class="language-html">&lt;meta property=&quot;og:title&quot; content=&quot;Know what&#39;s in the roastery, bag by bag&quot; /&gt;
&lt;meta
  property=&quot;og:description&quot;
  content=&quot;Stock, roast dates and reorders for small coffee roasters. Free for one roaster.&quot;
/&gt;
&lt;meta property=&quot;og:image&quot; content=&quot;https://brewlog.example/share/brewlog-v2.png&quot; /&gt;
&lt;meta property=&quot;og:image:width&quot; content=&quot;1200&quot; /&gt;
&lt;meta property=&quot;og:image:height&quot; content=&quot;630&quot; /&gt;
&lt;meta
  property=&quot;og:image:alt&quot;
  content=&quot;Brewlog&#39;s stock screen, four coffees listed with their roast dates&quot;
/&gt;
&lt;meta property=&quot;og:url&quot; content=&quot;https://brewlog.example/&quot; /&gt;
&lt;meta property=&quot;og:type&quot; content=&quot;website&quot; /&gt;
&lt;meta name=&quot;twitter:card&quot; content=&quot;summary_large_image&quot; /&gt;
</code></pre>
<p>X has its own <code>twitter:</code> tags, but in its markup reference each one falls back to its Open Graph twin, so <code>twitter:image</code> borrows <code>og:image</code> and <code>twitter:title</code> borrows <code>og:title</code>. The one you can&#39;t leave out is <code>twitter:card</code>. Without it, X says a small summary card &quot;may be rendered&quot; in place of the big picture, which is why <code>summary_large_image</code> sits at the bottom of that block.</p>
<p>Slack reads &quot;oEmbed and Twitter Card / Open Graph tags&quot;. Apple&#39;s 2017 talk on rich links says Messages uses <code>og:title</code> in place of your <code>&lt;title&gt;</code>, takes <code>og:image</code> for the picture, and keeps your site&#39;s icon (the <code>link rel=&quot;icon&quot;</code> one) as a fallback for when the connection is poor. If two plugins both write an <code>og:image</code>, ogp.me says the first one from the top wins, and that&#39;s often not the one you meant.</p>
<h2>The picture size each platform asks for</h2>
<p>One picture of 1200 by 630 pixels, saved as a PNG or JPEG under 600KB, fits every size these platforms publish (X&#39;s 2:1 near enough). Here&#39;s what each says, as of 17 November 2026.</p>
<table>
<thead>
<tr>
<th>Where</th>
<th>Size and shape</th>
<th>File size</th>
</tr>
</thead>
<tbody><tr>
<td>Facebook</td>
<td>At least 1200 by 630, close to 1.91:1 (200 by 200 at the very least)</td>
<td>8MB at most</td>
</tr>
<tr>
<td>LinkedIn</td>
<td>At least 1200 by 627, 1.91:1; under 401 wide shows as a thumbnail</td>
<td>5MB at most</td>
</tr>
<tr>
<td>X</td>
<td>2:1, from 300 by 157 up to 4096 by 4096; JPG, PNG, WEBP or GIF, no SVG</td>
<td>Under 5MB</td>
</tr>
<tr>
<td>WhatsApp</td>
<td>At least 300 wide, no wider than 4:1</td>
<td>Under 600KB</td>
</tr>
<tr>
<td>iMessage</td>
<td>No size given; a picture of this page rather than your logo</td>
<td>10MB with the icon and any video</td>
</tr>
<tr>
<td>Slack</td>
<td>No size given</td>
<td>No limit given</td>
</tr>
</tbody></table>
<p>Two notes on that table. X&#39;s own card pages have gone from its developer site (its new documentation has no page on cards that I could find on 17 November), so its row comes from the last copies, archived in January and February 2026. And WhatsApp&#39;s figures are from Meta&#39;s guide for its business platform, which says previews show &quot;when the link is sent via chat or shared via status&quot; without limiting that to business accounts.</p>
<p>A 1200 by 630 picture is a touch narrower than X&#39;s 2:1, so X may trim a sliver off the top or bottom. Its archived pages don&#39;t say how it trims, so I&#39;d keep anything that matters away from the edges. Apple&#39;s advice is that &quot;images should not include text&quot;, because rich links are drawn at so many sizes. If your card carries a headline anyway, as plenty of landing pages&#39; cards do, keep it to a few words in type big enough to read at thumbnail size on a phone.</p>
<p>Add <code>og:image:width</code> and <code>og:image:height</code> too. Facebook says that without them &quot;the first person who shares a piece of content won&#39;t see a rendered image&quot;, since its crawler has to download and process the picture first. With them, it can draw the card at once.</p>
<h2>Why a shared link shows no picture</h2>
<p>A blank card nearly always comes down to the app not getting your tags or not getting your picture. Here are the causes I&#39;d check, roughly in this order, most of them named in the platforms&#39; own documentation.</p>
<p>The tags are written by scripts. Apple says &quot;JavaScript does not run when creating rich links&quot;, so the tags &quot;need to be in the source of the page&quot;, and Slack fetches &quot;as little of the page as it can&quot;. A single-page app that adds its tags with a head helper after loading looks fine in your browser&#39;s inspector and empty to every one of these apps, the same trap that keeps a page from being <a href="https://magpieaudit.com/articles/named-in-ai-answers">named in AI answers</a>. View the page&#39;s source (not the inspector) and search it for <code>og:image</code>.</p>
<p>The picture&#39;s address doesn&#39;t work from outside. WhatsApp wants &quot;an absolute URL&quot;, so <code>/share.png</code> on its own may give it nothing. A build that bakes in <code>http://localhost:3000/og.png</code> or a staging address works on the developer&#39;s machine and nowhere else. LinkedIn adds that a picture on &quot;a protected directory or website&quot; won&#39;t show, and X&#39;s old troubleshooting page lists a robots.txt that turns away Twitterbot. Slack, for what it&#39;s worth, doesn&#39;t read robots.txt at all.</p>
<p>The file is too big or the wrong kind. 600KB is WhatsApp&#39;s line, and X doesn&#39;t take SVG. WhatsApp also wants your <code>&lt;head&gt;</code> within &quot;the first 300KB of the HTML&quot;, so a head stuffed with inlined scripts and styles can push the tags past where it stops reading.</p>
<h2>Why it shows the wrong picture or an old one</h2>
<p>An old card is a cache. Each platform keeps what it fetched, and keeps the picture by its address, so changing the file on your server while keeping its name changes nothing for anyone. Facebook says images &quot;won&#39;t be updated unless the URL changes&quot;. X&#39;s troubleshooting page said its crawler re-reads a page&#39;s tags &quot;roughly every seven days&quot; and caches pictures by address too. Slack holds its answer &quot;for around 30 minutes&quot;, the shortest any of them states. LinkedIn&#39;s help says a preview &quot;may show an old image that is cached&quot;, usually because the link was shared before.</p>
<p>So the card keeps showing last month&#39;s picture until you give the new one a new name. <code>brewlog-v2.png</code>, or the old name with <code>?v=2</code> on the end, which is the workaround X&#39;s old page suggested.</p>
<p>The wrong picture altogether is usually a leftover. Site builders fill in their own stock description and picture until you write yours, and a theme or SEO plugin can add a second <code>og:image</code> above yours. Either way a stranger has taken over your bird feeder, with the builder&#39;s logo or someone&#39;s stock photo sitting where your product should be.</p>
<h2>How to refresh a preview</h2>
<p>Change the picture&#39;s address first, publish, then ask each platform to read the page again. Cards already posted keep the old picture; LinkedIn says its refresh changes the preview &quot;for new posts that include it&quot;, and I&#39;d assume the same of the others.</p>
<table>
<thead>
<tr>
<th>Where</th>
<th>How to refresh</th>
</tr>
</thead>
<tbody><tr>
<td>Facebook</td>
<td>Paste the link into the Sharing Debugger, which &quot;also triggers a scrape&quot;</td>
</tr>
<tr>
<td>LinkedIn</td>
<td>Paste it into Post Inspector at linkedin.com/post-inspector</td>
</tr>
<tr>
<td>X</td>
<td>No current tool documented; a new image address, or wait about seven days</td>
</tr>
<tr>
<td>Slack</td>
<td>Wait about 30 minutes</td>
</tr>
<tr>
<td>WhatsApp, iMessage</td>
<td>No tool documented; a new image address is the best bet I know of</td>
</tr>
</tbody></table>
<p>WhatsApp&#39;s and iMessage&#39;s rows are my reasoning, not their documentation, which says nothing about caching. A new address is what Facebook and X say works for pictures, and it costs you nothing to try.</p>
<h2>What Magpie&#39;s free report checks</h2>
<p>Magpie&#39;s free report has a section called &quot;Your link, shared&quot; that reads your tags from the HTML your server sends, as these apps do, and draws the card your link would get on X, Slack, Facebook, LinkedIn, WhatsApp, iMessage and five more. It checks that a share picture is named and actually loads (and says so when its address points somewhere like localhost), that X is asked for the large card, that there&#39;s a share title and a description, and that none of it is a site builder&#39;s stock words or picture.</p>
<h2>Questions people ask</h2>
<h3>Should the share title match the page&#39;s title?</h3>
<p>Not word for word. The <code>&lt;title&gt;</code> is written for <a href="https://magpieaudit.com/articles/google-title-and-description">a Google result</a>, with your site&#39;s name after it. The share title shows with your domain under it, so Apple suggests leaving the site&#39;s name out of <code>og:title</code> altogether, and X&#39;s reference gave <code>twitter:title</code> 70 characters at most.</p>
<h3>Does every page need its own picture?</h3>
<p>The platforms would like it. X&#39;s card page asked for a picture &quot;representing the content of the page&quot; and not &quot;a generic image such as your website logo&quot;, and Apple says much the same, with a logo better used as the icon. On a small site I wouldn&#39;t lose sleep over it, though. Give the home page and each main landing page a picture of the product doing its job (the same thing your <a href="https://magpieaudit.com/articles/landing-page-first-screen">first screen</a> shows, ideally), and let blog posts draw theirs from their titles, which is what our own articles do.</p>
]]></content:encoded></item>
<item><title>Do your site&apos;s promises match your terms?</title><link>https://magpieaudit.com/articles/site-promises-and-your-terms</link><guid isPermaLink="true">https://magpieaudit.com/articles/site-promises-and-your-terms</guid><pubDate>Tue, 17 Nov 2026 22:30:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Pricing</category><category>Social proof</category><description>They have to agree. &quot;Cancel any time&quot;, &quot;we never share your data&quot;, a money-back guarantee and &quot;no hidden fees&quot; are promises, and UK and US law judge the impression your site gives, so a clause in your terms that quietly takes one back doesn&apos;t fix it. Read each promise against your terms and privacy policy, then change whichever side is wrong.</description><content:encoded><![CDATA[<p>Your site and your terms were probably written months apart, by different people or by the same person in two very different moods. The landing page got rewritten in launch week. The terms came from a template, or from a friend&#39;s startup, and nobody has opened them since.</p>
<p>So the page says &quot;Cancel any time&quot; and the terms say an annual plan runs to the end of its year. The page says &quot;We never share your data&quot; and the privacy policy lists four advertising partners. Each one reads fine on its own, and together they don&#39;t hold.</p>
<p>I&#39;m not a lawyer and none of this is legal advice. What follows is what the regulators&#39; own documents say, checked on 17 November 2026, and how I&#39;d go through a site to find the gaps before a buyer does.</p>
<h2>Why a clause in your terms doesn&#39;t fix a promise on your site</h2>
<p>Regulators in both countries look at the impression the whole site gives, and a promise in large type isn&#39;t cancelled by a clause most buyers will never open.</p>
<p>In the US, section 5 of the FTC Act declares &quot;unfair or deceptive acts or practices in or affecting commerce&quot; unlawful. The Federal Trade Commission&#39;s Policy Statement on Deception, a letter to Representative John Dingell dated 14 October 1983, says what that means. It finds deception where something is &quot;likely to mislead the consumer acting reasonably in the circumstances, to the consumer&#39;s detriment&quot;. It judges &quot;the net impression&quot;, &quot;the entire mosaic, rather than each tile separately&quot;. And it says plainly that &quot;written disclosures or fine print may be insufficient to correct a misleading representation.&quot; Express claims, the words you chose to put on the page, are presumed to matter to buyers.</p>
<p>In the UK the same idea sits in the Digital Markets, Competition and Consumers Act 2024, whose unfair commercial practices part came into force on 6 April 2025. Section 226 counts false or misleading information as a misleading action, and adds that &quot;an overall presentation may be deceiving even if the information it contains is true.&quot; Section 227 says leaving out material information includes giving it &quot;in such a way that the consumer is unlikely to see it&quot;.</p>
<p>The Competition and Markets Authority&#39;s guidance (CMA207, updated 18 November 2025) has an example that reads a lot like a pricing page. A shop advertises big discounts on televisions, and &quot;a footnote in very small font explains that only a very small number of the televisions are on sale&quot;. The advert may be factually correct, the CMA says, and if the average buyer wouldn&#39;t notice the footnote and would have decided differently with it, they&#39;ve been misled. Swap the footnote for clause 14.3 of your terms and you&#39;ve got the same problem.</p>
<h2>The promises that come apart most often</h2>
<p>Four lines turn up on a lot of software sites, and each has its own way of drifting away from the documents behind it.</p>
<h3>&quot;Cancel any time&quot;</h3>
<p>I&#39;d guess this is the line most often pinched, magpie fashion, from a rival&#39;s pricing page without a look at your own billing. Read your terms for a minimum term, a notice period, cancelling by email only, or a yearly plan that can&#39;t be ended early. If any of those are there, the line on the page is wrong, or at least incomplete, and it often sits right under the button, where it does the most persuading.</p>
<p>For trials, the UK&#39;s subscription rules (drafted to start on 1 January 2027) add their own list of what must be said before a buyer commits, which I&#39;ve gone through in <a href="https://magpieaudit.com/articles/free-trial-card-up-front">the article on free trials that take a card</a>.</p>
<h3>&quot;We never share your data&quot;</h3>
<p>Privacy promises are where I&#39;d look first. When our full report read the AI products that led a launch site in October 2026 against their own privacy policies and terms, six of the nine contradictions it found were about data, meaning what was stored, for how long, where, and who else got to see it.</p>
<p>The FTC case people tend to point to is BetterHelp. On 2 March 2023 the FTC said the online counselling service had &quot;promised consumers that it would not use or disclose their personal health data except for limited purposes&quot;, at several points in its sign-up, and had then passed data to Facebook, Snapchat, Criteo and Pinterest for advertising. Under the proposed order it was to pay $7.8 million to consumers and be banned from sharing health data for advertising.</p>
<p>Your privacy policy has to say who you share data with anyway. The ICO&#39;s guidance on the right to be informed lists &quot;who it will be shared with&quot; among the things people must be told, wants it &quot;concise, transparent, intelligible, easily accessible&quot;, and says &quot;you must regularly review, and where necessary, update your privacy information.&quot; My guess is that the policy is usually the more honest document, and the words on your site are what need softening. If an analytics script or an ad pixel sends your visitors&#39; data to another company, that company is who the policy should be naming, so the trackers <a href="https://magpieaudit.com/articles/cookie-banners-on-landing-pages">your cookie banner</a> answers for belong in the same check.</p>
<h3>&quot;30-day money-back guarantee&quot;</h3>
<p>The US guides on guarantees are short and blunt. The words &quot;Money Back Guarantee&quot; should be used &quot;only if the seller or manufacturer, as the case may be, refunds the full purchase price of the advertised product at the purchaser&#39;s request&quot;, and any &quot;material limitations or conditions&quot; have to be disclosed so prospective buyers notice and understand them (16 CFR 239.3). A refund &quot;at our sole discretion&quot; in your terms, or one that excludes the first month, doesn&#39;t sit with a guarantee badge on the pricing table.</p>
<p>In the UK there&#39;s a trap going the other way. Schedule 20 of the 2024 Act bans &quot;presenting rights given to consumers by law as a distinctive feature of the trader&#39;s offer&quot;. The CMA&#39;s example is a shop boasting of a 14-day no-questions refund that buyers already have under the cancellation rules. If your &quot;14-day money-back guarantee&quot; is the legal cooling-off period with a badge on it, I&#39;d either make the guarantee longer than the law gives or stop presenting it as a perk. Banned practices can carry penalties of up to £300,000 or 10% of worldwide turnover, whichever is higher, according to the same guidance.</p>
<h3>&quot;No hidden fees&quot;</h3>
<p>The fee doesn&#39;t have to be hidden in a sinister way to break this one. A card surcharge in the terms, a set-up fee that appears at checkout, VAT added for consumers after the price was shown. Section 230 of the UK Act says &quot;the total price of a product includes any fees, taxes, charges or other payments that the consumer will necessarily incur&quot;, and the same Act lists the total price among the material information an invitation to buy has to give. <a href="https://magpieaudit.com/articles/show-prices-on-landing-page">How to show prices honestly</a>, fees and yearly billing included, has its own article.</p>
<h2>Hardly anyone reads the small print, and it still matters</h2>
<p>Very few buyers read terms. Yannis Bakos, Florencia Marotta-Wurgler and David Trossen tracked 48,154 monthly visitors to 90 software companies&#39; sites and found that only one or two in every 1,000 shoppers opened the licence agreement at all, and most of those who did read only a small part of it (Journal of Legal Studies, 2014).</p>
<p>That cuts both ways. It&#39;s exactly why the FTC and CMA won&#39;t let the small print do the work of correcting the headline, because almost nobody gets that far. My guess is that the one or two in a thousand who do read it are the careful ones about to sign up for the biggest plan, or the ones who&#39;ve just been charged and are hunting for the cancel clause. That&#39;s nobody&#39;s study, only how it looks from here. Either way, those are the people who&#39;ll sing like canaries about a broken promise in a review, and I&#39;d bet louder than a happy customer mentions a kept one.</p>
<p>There&#39;s a newer reader too, and I&#39;m less sure how much weight to give it. People now ask AI assistants whether a product is safe to sign up for, and an assistant that can read your privacy policy can quote it straight back at them.</p>
<h2>How to check your own site in an afternoon</h2>
<p>Copy out every promise your site makes, then find the clause that backs each one. For most sites it&#39;s a slow afternoon, I&#39;d guess, and it goes a lot faster with the terms and the page open side by side.</p>
<ol>
<li>Collect the promises from the first screen, the line under each button, the pricing table and its footnotes, the FAQ and the footer. Badges count (&quot;GDPR compliant&quot;, &quot;Bank-level security&quot;).</li>
<li>For each one, search the terms and the privacy policy for its subject: cancel, refund, share, sell, retain, fee, charge.</li>
<li>Mark each promise as backed, contradicted, or not mentioned at all.</li>
<li>Fix whichever side is wrong. Sometimes that&#39;s the page (the policy is accurate, the marketing got ahead of it), sometimes the terms (a template clause you never meant).</li>
</ol>
<p>Here&#39;s how that might look for a made-up note-taking app at minutes.example:</p>
<table>
<thead>
<tr>
<th>On the site</th>
<th>In its terms or policy</th>
<th>What I&#39;d change</th>
</tr>
</thead>
<tbody><tr>
<td>Cancel any time</td>
<td>Annual plans run to the end of the term and are not refunded</td>
<td>&quot;Monthly plans cancel any time; yearly plans run to their renewal date&quot;</td>
</tr>
<tr>
<td>We never share your data</td>
<td>We share usage data with analytics and advertising partners</td>
<td>Drop the ad partner, or say &quot;We never sell your data&quot; and name the analytics provider</td>
</tr>
<tr>
<td>30-day money-back guarantee</td>
<td>Refunds are at our discretion</td>
<td>Rewrite the refund clause to give the full price back within 30 days on request</td>
</tr>
<tr>
<td>No hidden fees</td>
<td>A 3% card processing fee is added at checkout</td>
<td>Put the fee in the price shown, or stop charging it</td>
</tr>
</tbody></table>
<p>The &quot;not mentioned&quot; ones are worth a second look. A promise about how long you keep recordings, with nothing in the policy about retention, usually means the policy is the template and nobody filled that part in.</p>
<p>Magpie&#39;s full report does this read too. It finds your privacy policy and terms from your own links, sets the site&#39;s promises against them, and keeps a quote from a document only where the exact words are there, linked to the place in the document where the clause sits. Ordinary terms, such as a right to end accounts for abuse, aren&#39;t treated as contradictions.</p>
<h2>Questions people ask</h2>
<h3>Is a template privacy policy fine?</h3>
<p>As a starting point, yes. The trouble is that a template describes some imaginary business, often one with more partners and more tracking than yours, or fewer. Go through it line by line and make it describe what your site actually runs, then check the page against that.</p>
<h3>Should I change my site or my terms?</h3>
<p>Change whichever one describes your business wrongly. If the policy lists an ad pixel you really do run, the promise on your site has to go, or the pixel does. I&#39;d lean towards fixing the practice where you can, honestly, because a promise like &quot;we never share your data&quot; sells well and you&#39;d keep it.</p>
<h3>Does this apply if I only sell to businesses?</h3>
<p>The UK rules quoted here protect consumers, so selling only to businesses takes you out of those. I&#39;d still fix the gap. A buyer at another company who reads &quot;cancel any time&quot; on your site and then finds a 12-month term in the contract isn&#39;t going to trust much else you say, and that&#39;s a poor start to any renewal conversation.</p>
]]></content:encoded></item>
<item><title>Broken links on landing pages: which lose visitors</title><link>https://magpieaudit.com/articles/broken-links-on-landing-pages</link><guid isPermaLink="true">https://magpieaudit.com/articles/broken-links-on-landing-pages</guid><pubDate>Tue, 17 Nov 2026 22:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Calls to action</category><description>The dead ends that lose visitors are the ones met while trying to act, such as a main button or menu link that opens a 404, a &quot;#&quot; link that does nothing, a jump to a removed section, or a link that lands on the home page. Press every link on a phone and a laptop, and redirect each old address to its nearest new page with a 301.</description><content:encoded><![CDATA[<p>Picture a pottery studio at kilnhouse.example (made up) the week after its relaunch. The new site looks lovely. Its old classes page lived at /classes and now lives at /workshops, but the &quot;Book a class&quot; button in the header still points at /classes, so everyone who presses it lands on a 404 that says &quot;Not Found&quot; and nothing else.</p>
<p>Nobody at the studio sees it, because they type the new address from memory. The bookings just go quiet.</p>
<p>Most broken links aren&#39;t that bad. A dead link to a 2019 blog post in the footer costs you very little. The ones I&#39;d fix today are those a visitor meets while they&#39;re trying to do the thing your site is for, which on a landing page means <a href="https://magpieaudit.com/articles/call-to-action-button-text">the main button</a>, the menu and anything on <a href="https://magpieaudit.com/articles/landing-page-first-screen">the first screen</a>.</p>
<h2>Which broken links actually lose visitors</h2>
<p>Six kinds of dead end, roughly from worst to mildest:</p>
<table>
<thead>
<tr>
<th>What the visitor presses</th>
<th>What they meet</th>
</tr>
</thead>
<tbody><tr>
<td>A link to a page that&#39;s gone</td>
<td>A 404 or 410 page, often with no way forward</td>
</tr>
<tr>
<td>A link to a domain that&#39;s lapsed</td>
<td>The browser saying the site can&#39;t be reached</td>
</tr>
<tr>
<td>A button or &quot;#&quot; link with no script</td>
<td>Nothing at all, so they press it again, then leave</td>
</tr>
<tr>
<td>A link to your own broken page</td>
<td>A server error, the same on a second try</td>
</tr>
<tr>
<td>A jump to a section that was removed</td>
<td>Nothing at all; the page stays where it is</td>
</tr>
<tr>
<td>A link to one of your own pages</td>
<td>Your home page, with no word of why</td>
</tr>
</tbody></table>
<p>The two that do nothing are the sneakiest, I think. A 404 at least tells the visitor something went wrong. A button that does nothing tells them nothing, so they can&#39;t tell whether it&#39;s your site, their signal or their thumb.</p>
<h3>Links that point at &quot;#&quot;</h3>
<p>A link written as <code>href=&quot;#&quot;</code> or <code>javascript:void(0)</code> relies on a script to do its job, and when the script hasn&#39;t loaded, has an error or was taken out in the redesign, the link goes nowhere. MDN calls these &quot;fake buttons&quot;, and says they misbehave when someone opens the link in a new tab or bookmarks it, or while JavaScript &quot;is loading, errors, or is disabled&quot;. Its advice is a real <code>&lt;button&gt;</code> for actions, and a link only &quot;for navigation to a real URL&quot;.</p>
<p>On a phone this tends to show up in the menu. A burger menu whose &quot;Pricing&quot; item was wired to an old script is dead on mobile and fine on the laptop where you tested it.</p>
<h3>A jump to a section that isn&#39;t there</h3>
<p>A link like <code>/#pricing</code> scrolls to the part of the page with <code>id=&quot;pricing&quot;</code>. Remove that section, or rename its id, and the HTML Standard&#39;s rule is that the fragment &quot;does not identify anything&quot;, so the browser simply doesn&#39;t scroll. No error, no 404. The page just sits there while the visitor wonders why &quot;See pricing&quot; did nothing.</p>
<p>Your server never hears about it, either. As MDN puts it, the fragment &quot;is not sent to the server when the URI is requested&quot;, so no server log or 404 report will ever count these. You only find them by pressing them.</p>
<h3>A link that sends people home</h3>
<p>Some sites answer every missing page by sending the visitor to the home page. It feels tidy, but the visitor wanted the classes page and is now looking at the hero they&#39;ve already read, with no clue the page they asked for is gone. A link that comes home like a homing pigeon every time is as dead as a 404, only quieter.</p>
<p>Google takes the same view. Its site move guide warns against sending many old addresses to &quot;one irrelevant single URL destination, such as the home page&quot;, because it &quot;can confuse users and might be treated as a soft 404 error&quot;.</p>
<h2>How to find them</h2>
<p>Press everything, on a phone and on a laptop, including the menu once it&#39;s opened on the phone. That&#39;s the honest answer, and it&#39;s dull enough that I doubt many people do it after every release.</p>
<p>A link checker helps with the ones that have an address. Jakob Nielsen&#39;s advice in 1998 was to &quot;run a link validator on your site at regular intervals&quot;, and it still holds. What a crawler of that kind can&#39;t do is press a button and see whether anything happens, or tell a jump to a missing section from a working one, because neither of those ever asks the server for anything.</p>
<p>For addresses people arrive at from outside (old adverts, emails, other sites&#39; links), Google Search Console&#39;s page indexing report lists the ones Google found under &quot;Not found (404)&quot; and &quot;Soft 404&quot;, the 404s from the past month only. Your server&#39;s logs show which missing addresses people actually request, and how often. Neither shows you a dead &quot;#&quot; button.</p>
<p>Magpie&#39;s full report does the pressing for you. It follows every link, button and form on both devices once, in a browser of its own, and counts only what a visitor would meet: a page that says it&#39;s not found, a site that doesn&#39;t exist, a server error from your own site twice in a row, a &quot;#&quot; or button that does nothing when pressed, a jump to a part the page doesn&#39;t have, and a link to your own page that lands on the home page. Anything it can&#39;t be sure of (a firewall that turns automated visits away, a slow answer, where a form would send its data) is taken as working, so a dead end it shows is one a visitor would hit. Each is photographed where it sits on the phone and the laptop.</p>
<h2>What to do with the old addresses after a redesign</h2>
<p>Give every old address a permanent redirect to its nearest new page, done on the server, before the new site goes live. Pages migrate in a redesign, and the redirect is how anyone holding the old address follows them to where they&#39;ve landed.</p>
<p>Google&#39;s guide to redirects says to use &quot;a permanent server-side redirect whenever possible&quot;, and names 301 and 308 as the codes that mean a page &quot;has permanently moved to a new location&quot;. Its crawler documentation calls 308 &quot;equivalent to 301&quot;. A temporary 302 or 307 keeps the old address in search results, which is fine for a page that&#39;s coming back and wrong for one that isn&#39;t.</p>
<p>For kilnhouse.example the map is short:</p>
<table>
<thead>
<tr>
<th>Old address</th>
<th>Send it to</th>
<th>How</th>
</tr>
</thead>
<tbody><tr>
<td>/classes</td>
<td>/workshops</td>
<td>301</td>
</tr>
<tr>
<td>/classes/wheel-101</td>
<td>/workshops/beginners-wheel</td>
<td>301</td>
</tr>
<tr>
<td>/gift-vouchers</td>
<td>/shop/vouchers</td>
<td>301</td>
</tr>
<tr>
<td>/summer-2024-camp</td>
<td>Nothing new; it&#39;s over</td>
<td>410, with help</td>
</tr>
<tr>
<td>/#timetable</td>
<td>/workshops#timetable</td>
<td>Fix the link</td>
</tr>
</tbody></table>
<p>Write it as a spreadsheet before you touch the server, one row for every old address that had visitors or links from other sites. A few rules from Google&#39;s site move guide are worth keeping beside it. Map each old address to its own new one, never everything to the home page. Point straight at the final page, since a chain of redirects is slower and Google suggests &quot;ideally no more than 3 and fewer than 5&quot; hops. And &quot;keep the redirects for as long as possible, generally at least 1 year&quot;.</p>
<p>The last row in that table is a jump, so no redirect can help it (the server never sees the <code>#timetable</code> part). Search your own templates for links to ids that no longer exist and change them by hand.</p>
<p>When a page is really gone with nothing to replace it, Google says to answer with a 404 or 410. Nielsen saw old content a little differently in 1998. He&#39;d rather keep the address alive and point it at a follow-up or the nearest current page. I&#39;d lean that way too for anything that was ever linked from an advert or a newsletter.</p>
<h2>What your 404 page should say</h2>
<p>Say plainly that the page isn&#39;t here, then offer the likeliest ways on. NN/g&#39;s 2023 error message guidelines single out the 404 as the web&#39;s most common error message and one that breaks their rule on plain words, and they ask every error to &quot;offer some potential remedies&quot;. Nielsen&#39;s 1998 list for 404 pages still reads well: a short, slightly apologetic line, links to the pages people most often meant, and a search box.</p>
<p>On a small marketing site I&#39;d keep it to one line and three links: the home page, the pricing page and the main thing you sell. For kilnhouse.example, that&#39;s &quot;This page has moved or gone. Most people looking for it want&quot; followed by Workshops, Gift vouchers and the timetable.</p>
<p>Make sure it answers with a real 404 status, too. A &quot;page not found&quot; message sent with a 200 (success) status is what Google calls a soft 404, and it leaves the dead address looking alive to search engines and to link checkers alike.</p>
<h2>Do broken links hurt your Google ranking?</h2>
<p>Not on their own. Google&#39;s Susan Moskwa wrote in 2011 that if some URLs on your site 404, &quot;this fact alone does not hurt you or count against you in Google&#39;s search results&quot;. A page that returns a 404 does drop out of the index, though, and with it whatever traffic it earned, which is why redirecting moved pages matters more than the 404 count itself. (What your pages show in the results once they&#39;re found is covered in <a href="https://magpieaudit.com/articles/google-title-and-description">your Google title and description</a>.)</p>
<h2>Questions people ask</h2>
<h3>Should I use a 301 or a 308?</h3>
<p>Either. Google treats them the same. Some servers and hosting platforms only offer one, and that&#39;s fine.</p>
<h3>How often should I check for broken links?</h3>
<p>After every release that touches the menu, the main button or page addresses, and I&#39;d say about once a month otherwise. I&#39;d add one more moment that&#39;s easy to forget, which is whenever a partner, a payment provider or a booking tool you link to changes its own addresses, because their redesign breaks your link without anything on your side changing at all, and you&#39;ll only hear about it when someone emails to say the &quot;Book now&quot; button on your pricing page goes to a page that doesn&#39;t exist.</p>
<h3>Do broken links to other sites matter?</h3>
<p>Less than your own, unless they&#39;re on the path to buying. A dead link to a press article in the footer is a small blemish. A dead link to your booking or checkout provider can cost you the sale itself.</p>
]]></content:encoded></item>
<item><title>Which questions should your landing page answer?</title><link>https://magpieaudit.com/articles/questions-your-landing-page-should-answer</link><guid isPermaLink="true">https://magpieaudit.com/articles/questions-your-landing-page-should-answer</guid><pubDate>Tue, 17 Nov 2026 22:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Calls to action</category><category>First screen</category><description>Answer the doubts that stop people acting: what it costs, who it&apos;s for, how long setup takes, whether they can cancel, whether it works with their tools and whether their data is safe. Put the commonest doubt in a short line by the button, the rest in the sections they belong to, and keep the FAQ for rarer questions from your inbox, sales calls and searches.</description><content:encoded><![CDATA[<p>The questions worth answering are the ones that stop someone pressing your button. Nearly always that&#39;s what it costs, who it&#39;s for, how long it takes to set up, whether they can get out again, whether it works with what they already use and whether it&#39;s safe to hand over their data. If your inbox keeps getting the same email, your site is leaving one of those open, and the FAQ at the bottom probably isn&#39;t where the answer should live.</p>
<h2>The six doubts buyers bring</h2>
<p>Most visitors arrive carrying some version of the same six doubts. The made-up rota app below, for cafés, at shiftly.example, shows each with the kind of line that settles it.</p>
<table>
<thead>
<tr>
<th>The doubt</th>
<th>What the visitor is really asking</th>
<th>A line that answers it on shiftly.example</th>
</tr>
</thead>
<tbody><tr>
<td>Price</td>
<td>Can I afford this, and is anything hidden</td>
<td>&quot;£12 a month per café, staff included&quot;</td>
</tr>
<tr>
<td>Who it&#39;s for</td>
<td>Is this made for a business like mine</td>
<td>&quot;For cafés and bakeries with 5 to 40 staff&quot;</td>
</tr>
<tr>
<td>Setup time</td>
<td>How much of my week does this eat</td>
<td>&quot;Import your staff from a spreadsheet in 10 minutes&quot;</td>
</tr>
<tr>
<td>Getting out</td>
<td>Am I stuck if it doesn&#39;t work</td>
<td>&quot;Monthly, cancel from your account in two clicks&quot;</td>
</tr>
<tr>
<td>Works with what I use</td>
<td>Will it talk to my till and my payroll</td>
<td>&quot;Sends hours to your payroll and reads your till&quot;</td>
</tr>
<tr>
<td>Safety</td>
<td>Is my staff&#39;s data safe with you</td>
<td>&quot;Data kept in the UK, and you can delete it any time&quot;</td>
</tr>
</tbody></table>
<p>The research backs this list pretty closely. When Aurora Harley at Nielsen Norman Group wrote up what makes a site credible in 2016, one of her four factors was up-front disclosure, being open about everything that touches the customer&#39;s experience (base costs, fees, returns, shipping). In the studies behind it, sites that left out basic information &quot;were almost immediately ruled out of consideration&quot;, and people went to the ones that were more open. Baymard Institute&#39;s 2025 review of software marketing sites found 33% gave no page for a particular industry, and warned that without one &quot;many users may assume a service won&#39;t meet their needs&quot;. It says integrations are &quot;a critical factor in decision-making&quot; too.</p>
<p>Safety and getting out show up even in shops. In Baymard&#39;s checkout surveys of US shoppers, 19% of those who gave up on an order said they didn&#39;t trust the site with their card details and 13% didn&#39;t like the returns policy. That&#39;s retail, so I&#39;d not lean on the figures for a software site, but the doubts carry over.</p>
<h2>Answer the biggest doubt next to the button</h2>
<p>Whatever doubt most of your buyers have goes right beside the main button, in a few words. That&#39;s the moment they&#39;re deciding, and a visitor who has to scroll off to an FAQ to learn whether there&#39;s a contract may not scroll back. I&#39;ve written about <a href="https://magpieaudit.com/articles/call-to-action-button-text">the line under the button</a> separately, so I&#39;ll only say it should hold one doubt, the commonest, and only what&#39;s true (whether <a href="https://magpieaudit.com/articles/free-trial-card-up-front">a free trial asks for a card</a> is a doubt in its own right).</p>
<p>Everything else belongs in the section a visitor reads when the doubt comes up. Price goes with the plans, and <a href="https://magpieaudit.com/articles/show-prices-on-landing-page">showing the price at all</a> settles more than any FAQ answer about it. Who it&#39;s for goes in the headline or the line under it. Integrations go with the features, as named logos or a list a buyer can search, and Baymard&#39;s testing found people look for an integrations link in the main navigation, the footer and on the home page. Safety sits near wherever you ask for data.</p>
<p>There&#39;s a reason not to leave a big answer behind a link. Raluca Budiu at Nielsen Norman Group described in 2020 how visitors who don&#39;t find enough to go on &quot;quickly decide that the page is not worth exploring any longer and simply leave.&quot; A link labelled &quot;FAQ&quot; smells of very little, and asks the buyer to bet a click that their question is inside.</p>
<h2>What the FAQ is for</h2>
<p>The FAQ is for the long tail, the questions a few buyers ask that the rest don&#39;t need in front of them. Refunds for a café that closes, student discounts, what happens to the rota if your Wi-Fi drops. Those are real, but they&#39;d clutter a section.</p>
<p>Two camps disagree about this, and I think both are right about different pages. Susan Farrell at Nielsen Norman Group argued in 2014 that FAQs still earn their place. People reading one &quot;may be judging your products, services, and company prior to purchasing&quot;, so the questions have to be real ones. &quot;Questions should echo the real concerns people contact you about,&quot; she wrote, in &quot;the vocabulary and phrasing of your visitors&quot;. The government content designers go the other way. Sarah Richards at the UK&#39;s Government Digital Service wrote in 2013 that &quot;if a question is frequently asked, it means you need that content on your site&quot;, structured so nothing needs repeating. Canada&#39;s digital team called FAQs &quot;a dumping ground&quot; for whatever a department wanted to push, and kept them in one place only, a short set of answers above the email address on its contact pages. On its general enquiries page, the share of visitors who went on to send an email fell from a high of 46% to an average of 20%.</p>
<p>So a question people ask a lot gets an answer in the page itself, and the FAQ holds the rest. If a big doubt has been pigeonholed down there because it was awkward to fit anywhere else (cancelling is the usual one), I&#39;d move it up.</p>
<h2>How to find the questions people really ask</h2>
<p>Read what buyers already send you, before you guess. Four places hold most of it.</p>
<ol>
<li>Your support inbox and chat, filtered to people who haven&#39;t paid yet. Sort by how often a question comes, and copy their wording, since that&#39;s how the next buyer will think it.</li>
<li>Sales and demo calls. Whatever someone asks in the first five minutes is what your site didn&#39;t tell them.</li>
<li>The searches that bring people to you. Google Search Console&#39;s Performance report groups your traffic &quot;by the search query users typed&quot;, and a query like &quot;shiftly cancel&quot; or &quot;shiftly payroll&quot; is a question asked straight into the search box.</li>
<li>Reviews of your rivals, the two and three star ones especially. They tell you which doubts turned out to be justified elsewhere, and so which ones your buyers will bring. <a href="https://magpieaudit.com/articles/comparing-yourself-with-rivals">Comparing yourself with rivals</a> has more on doing that fairly.</li>
</ol>
<p>One swallow doesn&#39;t make a summer, and one odd email doesn&#39;t make a frequently asked question. I&#39;d want to see a doubt come up several times before it gets a line on the page, and I&#39;d count it again after a change, because the questions that keep coming are the ones your answer didn&#39;t settle.</p>
<p>Watch for the answers your support team ends up typing again and again. If someone is copying the same paragraph into replies every week, that paragraph is half written already and only needs a place on the page.</p>
<h2>FAQ markup won&#39;t win you a search result now</h2>
<p>Google narrowed FAQ rich results in August 2023 to &quot;well-known, authoritative government and health websites&quot;, and its documentation said in 2026 that the feature would no longer appear in Google Search from 7 May that year. John Mueller&#39;s 2023 post says there&#39;s &quot;no need to proactively remove&quot; the markup, so leave it if it&#39;s there. Just don&#39;t write an FAQ for the sake of the snippet. Write it for the person reading, who is weighing whether to trust you.</p>
<h2>What Magpie&#39;s full report checks</h2>
<p>Our full report has a section called &quot;Questions your page leaves open&quot;. It reads the whole page as a buyer of its kind would and lists the questions they&#39;d still have, then checks each one against every section and drops any the page already answers, counting a short line under a button as much as a whole section. It shows four, each with the kind of visitor who&#39;d ask it. Beside it, three likely buyers read the page and say what would stop them and what would win them over, so you can see whether the doubt you think matters is the one they raise.</p>
<h2>Questions people ask</h2>
<h3>Should the FAQ be on the landing page or a page of its own?</h3>
<p>A handful of questions can sit near the foot of the landing page, where someone who scrolled that far is still weighing it up. Farrell&#39;s advice for a long FAQ or long answers was a list of questions leading to their own pages, which is roughly how a help centre works, and I&#39;d keep the two apart. The landing page&#39;s few should be about buying. How to export a rota belongs in the help centre.</p>
<h3>Does a chat widget do the same job?</h3>
<p>It catches questions, which is useful, but a buyer who has to ask has already had to stop. I&#39;d treat what people type into it as a list of answers your site is missing, and add them to the page, so the next buyer reads the answer without waiting for someone to parrot it back.</p>
]]></content:encoded></item>
<item><title>Comparing your product with rivals on your site</title><link>https://magpieaudit.com/articles/comparing-yourself-with-rivals</link><guid isPermaLink="true">https://magpieaudit.com/articles/comparing-yourself-with-rivals</guid><pubDate>Tue, 17 Nov 2026 21:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Pricing</category><category>Social proof</category><description>Yes, you can. UK, EU and US rules all allow a comparison with a named rival when it&apos;s true and like for like: the same kind of product, features buyers care about, figures as they stand today, and a way to check each one. &quot;Cheapest&quot;, &quot;#1&quot; and &quot;the only&quot; are read as claims about the whole market, so they need evidence covering all of it.</description><content:encoded><![CDATA[<p>Take a made-up staff rota app at rotafy.example. Its pricing page has a table with Rotafy and two named rivals across the top, twelve rows down the side, and a tick in every one of Rotafy&#39;s cells. The rivals get four ticks each and eight blanks. It looks like a clean win, and it&#39;s very close to the shape the UK&#39;s advertising regulator has ruled against twice since 2023, both times after a rival complained.</p>
<p>So yes, you can compare yourself with rivals by name, on your own site, in the UK, the EU and the US. What gets a comparison pulled is a buyer being unable to check it, and the people most likely to read your table line by line are the rivals in it.</p>
<h2>Naming a rival is allowed, on conditions</h2>
<p>In the UK the conditions sit in regulation 4 of the <a href="https://www.legislation.gov.uk/uksi/2008/1276/regulation/4">Business Protection from Misleading Marketing Regulations 2008</a>, last amended on 6 April 2025. A comparison has to set products side by side that meet &quot;the same needs or intended for the same purpose&quot;, and compare &quot;material, relevant, verifiable and representative features&quot; of them, price included. It mustn&#39;t confuse anyone about whose product is whose, denigrate the rival, or take unfair advantage of the reputation of its trade mark. The EU&#39;s Directive 2006/114/EC uses the same words in Article 4, and the UK&#39;s advertising code repeats them as rules 3.33 to 3.43.</p>
<p>That code has covered what a company says on its own website since 2011, so a comparison page is held to it as an advert would be. I&#39;m not a lawyer, and none of this is legal advice, but the rules and rulings below are short enough to read for yourself.</p>
<p>The US is warmer about the whole thing. The Federal Trade Commission&#39;s 1979 policy statement (16 CFR 14.15) &quot;encourages the naming of, or reference to&quot; competitors, &quot;but requires clarity, and, if necessary, disclosure to avoid deception of the consumer&quot;. It goes further than the UK in one place, saying &quot;disparaging advertising is permissible so long as it is truthful and not deceptive&quot;, where the UK regulation lists not discrediting or denigrating a rival as a condition of its own, beside being true.</p>
<p>Whether a comparison is worth making at all is a separate question, and the evidence is old and a bit thin. Dhruv Grewal and colleagues pooled studies of comparative adverts in 1997 and found they won more attention, better memory of the brand and stronger intentions to buy than adverts comparing with no one. People also believed the advertiser less and liked the advert less, and new brands comparing themselves with established ones seemed to gain most. That&#39;s adverts rather than landing pages, so I&#39;d take one thing from it. A comparison costs you some trust as it lands, and facts a buyer can check are the only way I know of to earn it back.</p>
<h2>What a fair comparison table holds</h2>
<p>The two rulings are worth reading in full, because they&#39;re the rota table above with the names changed. On 14 June 2023 the Advertising Standards Authority ruled against a blog post on a company&#39;s own site that compared its CRM tool with a named rival and one unnamed one. Every feature was ticked for the advertiser and far fewer for the others. The ASA said the rival&#39;s unticked features would be read as &quot;not available&quot;, and the advertiser &quot;did not provide any evidence in relation to RoboReception&#39;s capabilities&quot;. A note at the top telling readers features might have changed and to check with each provider didn&#39;t save it.</p>
<p>On 10 June 2026 it was an estate agent&#39;s leaflet and Facebook post, nine rows, all nine ticked for the agent and two for each of three rivals. Some rows were things the rivals offered too (&quot;the competitors also offered personalised sale boards&quot;), and &quot;Skilled negotiations by an agent who knows your home&quot; was a matter of opinion dressed as a feature. The table gave no sources and no date either, which was a breach on its own.</p>
<p>Both tables made the same mistake. A blank cell under a rival&#39;s name says &quot;they don&#39;t have it&quot;, and not finding something on a rival&#39;s site doesn&#39;t mean the rival lacks it. A row you couldn&#39;t check is a row to leave out, or one to fill with what you did find and where.</p>
<p>Choose the rows buyers weigh, rather than the ones that happen to flatter you. A table you win twelve rows to nil has been built to feather your own nest, and I think most readers can tell. Kate Moran and Taylor Dykes at Nielsen Norman Group wrote in 2024 that inconsistent content is &quot;the biggest problem with most comparison tables&quot;, and advised five items at most, with every row labelled, since a tick can&#39;t stand alone. That&#39;s expert review rather than a test, but it matches what the rulings punish.</p>
<p>Here&#39;s the rotafy.example table rebuilt, against two made-up rivals at shiftgrid.example and teamrota.example:</p>
<table>
<thead>
<tr>
<th></th>
<th>Rotafy</th>
<th>ShiftGrid</th>
<th>TeamRota</th>
</tr>
</thead>
<tbody><tr>
<td>Price for 20 staff, billed monthly</td>
<td>£60 a month</td>
<td>£80 a month</td>
<td>£45 a month, up to 25 staff</td>
</tr>
<tr>
<td>Shift swaps from the phone app</td>
<td>Yes</td>
<td>Yes</td>
<td>Yes</td>
</tr>
<tr>
<td>Payroll export</td>
<td>Xero and Sage</td>
<td>Xero</td>
<td>Not listed on its site</td>
</tr>
<tr>
<td>Free plan</td>
<td>Up to 5 staff</td>
<td>None listed</td>
<td>Up to 10 staff</td>
</tr>
</tbody></table>
<p>Under it, one line says each figure comes from that product&#39;s own pricing page on 3 November 2026, with a link to each. TeamRota is cheaper, and the table says so. Prices are compared at one team size and one billing period, the same care <a href="https://magpieaudit.com/articles/show-prices-on-landing-page">showing your own prices</a> asks for, since the UK code&#39;s rule 3.38 wants the basis of any price comparison made clear.</p>
<h2>Make every figure checkable</h2>
<p>The ASA&#39;s guidance on verifiability says a comparison should come with &quot;a signpost in the ad to information about the basis of the comparison&quot;, and that &quot;the information must also be readily accessible&quot;. Results behind a paywall didn&#39;t count in a 2020 ruling, and a 2022 footnote at the bottom of the page wasn&#39;t a clear enough signpost. It also advises against making a comparison at all when you can&#39;t make its data public.</p>
<p>On a website that&#39;s easy enough. Date the table, link each rival&#39;s figure to the page you took it from, and say in a sentence how you compared (which plan, what team size, monthly or yearly).</p>
<p>Then keep it true. The ASA&#39;s guidance on comparisons with named competitors says &quot;marketers must hold up to date evidence&quot;, and a rival&#39;s pricing page can change on a Tuesday afternoon without telling you. I&#39;d put a monthly reminder in the diary to check every figure in the table, and take the table down if nobody&#39;s going to.</p>
<h2>&quot;Cheapest&quot;, &quot;#1&quot; and &quot;the only&quot; mean the whole market</h2>
<p>An unqualified superlative is read as a claim against every rival you have, so the evidence behind it has to cover all of them as well.</p>
<p><strong>Cheapest.</strong> On 16 November 2016 the ASA ruled against a parcel broker&#39;s &quot;The Cheapest Parcel Delivery in the UK&quot;, which buyers would take to mean it &quot;would always have a cheaper option currently available&quot; than every other parcel service in the UK. Its price match didn&#39;t help. The ASA&#39;s guidance on lowest price claims (updated 10 November 2025) says they need evidence that you&#39;ll &quot;always beat, and not merely match&quot; rivals&#39; prices.</p>
<p><strong>#1.</strong> The ASA reads an unqualified &quot;number 1&quot; much as it reads &quot;best-selling&quot; or &quot;leading&quot;, and &quot;absolute &#39;number 1&#39; claims are likely to be understood as entire market comparisons&quot;, which usually means unit sales or market share for the whole market. In July 2022 a breakdown service&#39;s &quot;The UK&#39;s number 1&quot; claim fell because it didn&#39;t have its competitors&#39; membership figures. A &quot;#1 rated&quot; claim that only holds after filtering a review site is its own trap, and I go through it in the one <a href="https://magpieaudit.com/articles/ratings-and-review-badges">about star ratings and review badges</a>, with the ASA&#39;s own example.</p>
<p><strong>The only.</strong> A product that truly is the only one with some feature is a rare bird, and you&#39;ll need the evidence that it is. On 27 February 2019 the ASA ruled against a broadband provider&#39;s &quot;only major provider to guarantee no mid-contract price rises&quot;, after another provider complained. There was no accepted definition of a &quot;major provider&quot;, and the complainant offered fixed prices too. In May 2023 an over-50s life insurer&#39;s &quot;only&quot; claim stood, because it showed that every rival&#39;s cover came with a waiting period of 12 to 36 months.</p>
<p>So I&#39;d swap the superlative for the measured fact behind it. &quot;£20 a month less than ShiftGrid for 20 staff, billed monthly, on 3 November 2026&quot; is a smaller claim than &quot;cheaper&quot;, and it&#39;s true, and anyone can check it.</p>
<h2>A &quot;vs&quot; page is a comparison too</h2>
<p>A page called &quot;Rotafy vs ShiftGrid&quot; compares you with an identifiable competitor, so every condition above applies to it, the trade mark ones included. Using a rival&#39;s name in plain words is how you identify them. Making your page look like theirs, or calling your product a &quot;ShiftGrid clone&quot;, runs into the conditions on confusion and on presenting a product &quot;as imitations or replicas&quot;, and I wouldn&#39;t go near either.</p>
<p>Keep the tone flat and factual. Denigrating a rival is a breach in the UK even where the US would let a true put-down stand, so I&#39;d stick to what a buyer could check for themselves.</p>
<p>And date the page. In my view a vs page goes stale faster than anything else on a site, because nobody looks at it again after launch week, and I&#39;d guess the rival in its title reads it more often than you do.</p>
<p>Before naming anyone, it&#39;s worth asking who you&#39;re really up against. April Dunford&#39;s positioning guide (2021) starts with &quot;What would a customer do if your offering didn&#39;t exist?&quot;, and the answer is often &quot;Do nothing&quot;, which &quot;could mean using a spreadsheet, using a manual process, or hiring an intern to do it.&quot; A table set against the spreadsheet your buyers use today needs no rival&#39;s facts at all.</p>
<h2>What Magpie checks</h2>
<p>Magpie&#39;s full report finds up to three rivals through web searches and takes each rival&#39;s price only from that rival&#39;s own site. It reads your plans beside those prices like with like, and anything it didn&#39;t find on a rival&#39;s site it treats as unknown, never as missing, which is the line both of those tick tables crossed. The posts it writes for you never name a rival, since a named comparison needs the checking above.</p>
<h2>Questions people ask</h2>
<h3>Can I compare with &quot;other tools&quot; without naming any?</h3>
<p>You can, though it isn&#39;t the safe harbour it looks. The UK code&#39;s rule 3.37 says a comparison with an unidentifiable competitor still mustn&#39;t mislead, and its elements mustn&#39;t be picked to give you an unrepresentative advantage. In July 2019 a firm&#39;s &quot;only&quot; claim, shown beside a &quot;Three Best&quot; logo, was read by the ASA as a claim against every relevant competitor, which the firm couldn&#39;t back up. So &quot;most rota apps charge per user&quot; needs about the same evidence as a named table, and gives the buyer less to check, which is why I&#39;d rather name the rivals and link their pages.</p>
<h3>Do I need a rival&#39;s permission to name them?</h3>
<p>No. None of the conditions in regulation 4 is the rival&#39;s consent. Their right is to complain, and both tick-table rulings started with a complaint from a competitor.</p>
]]></content:encoded></item>
<item><title>Should your landing page show customer logos?</title><link>https://magpieaudit.com/articles/customer-logos-on-landing-pages</link><guid isPermaLink="true">https://magpieaudit.com/articles/customer-logos-on-landing-pages</guid><pubDate>Tue, 17 Nov 2026 21:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Social proof</category><description>Show a company&apos;s logo only when it really uses your product, has said yes in writing, and is a name your buyers know or a business like theirs. Keep &quot;As seen in&quot; for recent editorial coverage, never a paid ad, and know that the UK&apos;s advertising regulator has ruled against logo strips with no permission behind them.</description><content:encoded><![CDATA[<p>Five grey marks under &quot;Trusted by&quot;. To a stranger that row says five real companies chose you, still use you and are happy for the world to know it, and nobody can check any of that on the spot, which is exactly why it gets padded so often.</p>
<p>I&#39;d put a logo on the strip only when the company is a customer who said yes, and when your buyer either knows the name or sees a business like their own in it. Each mark is a claim about one company, so it has to be true of every one of them.</p>
<h2>What the evidence says about logo strips</h2>
<p>Not much, and less than the blog posts would have you think. The figure that gets repeated most, that client logos lift conversions by nearly 70%, goes back to a 2013 post on a testing tool&#39;s blog, and that post now redirects to a 404. I wouldn&#39;t build a page on it.</p>
<p>The closest thing to a real experiment I could open is small. In 2020 Lars Schmitt, Isabel Haupenthal and Faisal Bin Ahmed at the Technical University of Munich built a site for a made-up security start-up and ran Google Ads to it for 90 days. Of the 456 visits, 238 saw three trust elements together (logos of a pretend network of providers, a testimonial, and LinkedIn buttons for the founders). Those visits clicked the request button 41 times and sent five real requests. The other 218 clicked 26 times and sent none.</p>
<p>Time on the site barely moved, 36 seconds against 31.</p>
<p>Five against none is a hint and nothing more. The authors say themselves they couldn&#39;t run a significance test, because their free tools wouldn&#39;t export the raw data, and since the three elements went in together nobody can say how much of it the logos did. My reading, for what it&#39;s worth, is that proof matters most at the moment someone decides whether to ask, and that&#39;s usually where a logo row sits.</p>
<h2>Which logos to show</h2>
<p>A café owner looking at a till app learns more from three independent coffee shops than from one airline. Starlings follow starlings, and buyers aren&#39;t so different. So show the logos your buyer will recognise or relate to, from customers who use you today, and don&#39;t lean on fame as much as you might expect to.</p>
<p>The US rules treat a logo as an endorsement. The FTC&#39;s Endorsement Guides count a company&#39;s name or seal as one, and an endorser said to use the product &quot;must have been a bona fide user of it at the time the endorsement was given&quot; (16 CFR 255.0 and 255.1). You can only keep running it while you&#39;ve good reason to think they still use it, so a customer who left in the spring comes off the strip in the spring.</p>
<p>I&#39;m strict about one thing here. &quot;Trusted by&quot; reads as the company chose you, and one person at a bank signing up for your free plan with a work email isn&#39;t the bank choosing you. I&#39;d leave that logo off until the bank says otherwise.</p>
<p>Agencies and freelancers have a trap all their own. In 2018 the UK&#39;s Advertising Standards Authority ruled against a PR agency whose &quot;Trusted by&quot; logos and quotes came from work a staff member had done at a previous employer, finding that businesses &quot;would not expect the logos and testimonials to refer to work undertaken by an employee&quot; somewhere else. If it wasn&#39;t your company&#39;s work, it isn&#39;t your company&#39;s logo to show.</p>
<h2>How many logos to put in the row</h2>
<p>I couldn&#39;t find a study that tests a number, so this bit is just my opinion. One row of five to eight does the job, every logo readable on a phone without zooming in, which on a narrow screen often means four with the rest a swipe away or cut.</p>
<p>I&#39;d skip the endless scrolling marquee. It hides most of the logos most of the time, and I&#39;ve never been sure what it proves beyond there being a lot of them.</p>
<p>Take ledgerly.example, a made-up bookkeeping app. Its strip scrolls 14 logos, a supermarket chain and a telecoms giant among them, there because someone on their staff tried the free plan once. Cut it to the six accounting firms that pay for it and agreed in writing, and the row gets shorter, but every name on it is now a peer of the visitor and every one would stand up if somebody asked.</p>
<h2>Permission to use a customer&#39;s logo</h2>
<p>Ask in writing before a logo goes up. Every time. The CAP Code, which UK ads have to follow, says an ad &quot;must not feature a testimonial without permission&quot; (rule 3.50) and that you must hold evidence an endorsement is genuine, with the contact details of whoever gave it (rule 3.47).</p>
<p>In the agency ruling above, the ASA read &quot;Trusted by&quot; as meaning the companies &quot;had allowed their logos and testimonials to be used&quot;, and told the agency to take down every logo it didn&#39;t have express permission for. The whole case started with one complaint, from a rival, which is a fairly ordinary way for a logo used without asking to get caught out, and then the strip comes down quicker than a startled wren. The ASA&#39;s advice on official endorsements is blunter still. &quot;Only use official logos, trademarks, or branding from third parties if you have explicit permission.&quot;</p>
<p>There&#39;s a separate question too, because the logo is the customer&#39;s trademark. Microsoft&#39;s brand guidelines are a fair sample of how a big company handles it. Anyone may use its name in text &quot;to truthfully convey information about your product or service&quot;, but &quot;Don&#39;t use Microsoft&#39;s logos, icons, or designs, in any manner&quot; without a licence, and &quot;Don&#39;t imply an affiliation, endorsement, sponsorship, or approval&quot;.</p>
<p>US courts draw a similar line. The Ninth Circuit&#39;s test for naming another company&#39;s mark allows &quot;only so much of the mark or marks&quot; as is &quot;reasonably necessary&quot;, and its example is that a rival may name Coca-Cola but &quot;would not be entitled to use Coca-Cola&#39;s distinctive lettering&quot; (Playboy v. Welles, 2002). I&#39;m not a lawyer and this isn&#39;t legal advice. Writing a customer&#39;s name is a smaller step than drawing their logo, but I wouldn&#39;t take either without asking.</p>
<p>The ask itself is one short email to your contact. Say which page the logo goes on, show them the row it&#39;ll sit in, and ask whether their marketing team needs to sign it off. Check your contract for a publicity clause first, as it may already settle the question one way or the other. Then keep the reply somewhere you&#39;ll find it again, since that&#39;s the evidence the CAP Code expects you to hold.</p>
<h2>When &quot;As seen in&quot; press logos mislead</h2>
<p>CAP&#39;s advice says the ASA is likely to read &quot;As seen in&quot; as a claim that a journalist chose to write about you. So an ad you paid for in that paper doesn&#39;t count (CAP suggests &quot;As advertised in&quot; when that&#39;s the truth), the product has to be recognisably yours in the coverage, and the coverage has to be recent.</p>
<p>The rulings behind that advice are specific. In 2012 a mattress seller&#39;s &quot;As seen in&quot; claims were ruled misleading for every title but one, the only one where it hadn&#39;t paid for the space. In 2013 a TV appearance from 2009 was judged too old to support &quot;As seen on TV&quot;.</p>
<p>And in July 2026 the ASA ruled against a hair styling company whose ads showed &quot;SEE ON&quot; over the logos of Marie Claire, GQ, Allure and Cosmopolitan, with Elle, Vogue, Allure and Harper&#39;s Bazaar on its product page. It found &quot;no evidence&quot; the endorsements were genuine.</p>
<table>
<thead>
<tr>
<th>What you have</th>
<th>What the strip can honestly say</th>
</tr>
</thead>
<tbody><tr>
<td>A journalist&#39;s article about you, this year</td>
<td>&quot;Featured in&quot;, linked to the article</td>
</tr>
<tr>
<td>An ad you paid for in that paper</td>
<td>&quot;As advertised in&quot;, or nothing</td>
</tr>
<tr>
<td>A press release a distribution service syndicated</td>
<td>Nothing, as I read it, since nobody chose to cover you</td>
</tr>
<tr>
<td>One line in a round-up three years ago</td>
<td>Move it to a press page with the date beside it</td>
</tr>
</tbody></table>
<p>That press release row is my own reasoning from the ASA&#39;s editorial test rather than a ruling I&#39;ve found, so take it as that. Paying a wire service to post your release on a news site&#39;s partner pages still looks to me like paying for the space.</p>
<p>Partner and integration logos need the same care. In 2023 the ASA ruled against a door maker that showed &quot;Schüco Partner&quot; and &quot;Cortizo UK Official Dealer&quot; logos with no evidence of either partnership, and UK law bans the wider move outright. Schedule 20 of the Digital Markets, Competition and Consumers Act 2024 lists claiming a business or product &quot;has been approved, endorsed or authorised by a public or private body&quot; when it hasn&#39;t. If your product connects to Slack, say &quot;Works with Slack&quot;, and keep &quot;Partners&quot; for companies that have agreed to be one.</p>
<h2>Label the strip with what is true</h2>
<p>&quot;Customers include&quot; says only that they pay you. &quot;Trusted by&quot; says they chose you and are glad to be named, &quot;Works with&quot; is for integrations and &quot;Featured in&quot; for journalism, so pick the heading that matches what you&#39;ve actually got.</p>
<p>Then link what you can. A press logo that opens the article lets a doubtful visitor check it in one click, for the same reason <a href="https://magpieaudit.com/articles/ratings-and-review-badges">a rating badge should link to its reviews</a>. A logo with a named quote from someone at that company beside it does more than either does alone, because <a href="https://magpieaudit.com/articles/landing-page-testimonials">a believable testimonial</a> puts a person behind the mark.</p>
<p>Place matters a bit too. I&#39;d put a single row just under <a href="https://magpieaudit.com/articles/landing-page-first-screen">the first screen&#39;s</a> main button, where it answers &quot;who else uses this&quot; at the moment a visitor is deciding whether to scroll on or press it.</p>
<h2>What Magpie checks</h2>
<p>Magpie&#39;s free check counts the quotes on your site and the pictures marked as logos. When it finds no quote and no row of three logos or more, it asks for one named quote or customer logo near the top. It can&#39;t tell whether a logo belongs to a paying customer, or whether you were ever given permission to show it, so that part stays with you.</p>
<h2>Questions people ask</h2>
<h3>What if a customer asks me to take their logo down?</h3>
<p>Take it down that day and thank them. Their permission was the only thing holding it up, and the FTC&#39;s guides tie an endorsement to someone you&#39;ve good reason to think is still a customer.</p>
<h3>Should the logos be grey or in their own colours?</h3>
<p>I&#39;d go grey, or one tone, mostly so a row of other people&#39;s brand colours doesn&#39;t pull <a href="https://magpieaudit.com/articles/where-visitors-look-first">the eyes</a> away from your own button. I haven&#39;t found a study on it either way, so it&#39;s a taste call more than anything. The catch is that the logo&#39;s still their trademark, and some brand teams are fussy about anyone recolouring it, so mention the grey version in your permission email and let them say yes to that exact thing rather than finding out later that they mind.</p>
]]></content:encoded></item>
<item><title>Should a free trial ask for a card up front?</title><link>https://magpieaudit.com/articles/free-trial-card-up-front</link><guid isPermaLink="true">https://magpieaudit.com/articles/free-trial-card-up-front</guid><pubDate>Tue, 17 Nov 2026 21:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Pricing</category><category>Calls to action</category><description>Ask for a card up front only when the product shows its worth within the trial and you&apos;ll say plainly, beside the button, what is charged after it, on which day and how to cancel. Seven days is a sound length for most software. UK rules drafted for 1 January 2027 also want a reminder before the first charge.</description><content:encoded><![CDATA[<p>A card up front makes sense when the product proves itself inside the trial and you&#39;re open about the charge that follows. Ask for one on a product that takes a month to understand, or tuck the charge into grey type under the button, and you&#39;ll get fewer sign-ups plus a list of people paying for something they forgot they started.</p>
<p>I went looking for a controlled test of card against no card and didn&#39;t find one. There are plenty of companies writing up their own funnels, but none of it is a study I&#39;d lean on.</p>
<p>What does exist is a big field test of trials that renew by themselves against trials that end by themselves, which is pretty much the choice a card up front forces.</p>
<h2>What a card up front does to sign-ups</h2>
<p>It cuts them. Klaus Miller, Navdeep Sahni and Avner Strulov-Shlain offered 1.4 million readers of a European newspaper a promotional subscription that either rolled into a paid plan unless cancelled or ended on its own unless renewed, and everyone gave payment details at the start in both versions (so the card itself wasn&#39;t the difference, only what happened by default). In their January 2026 paper, offering the auto-renewing version &quot;lowers subscriptions take-up by 35% and total subscribers by 23% over 20 months&quot;.</p>
<p>For a few months after the promotion the renewing version had more people paying, and its cumulative revenue ran about 2.5 cents higher for each reader offered it, a gap that was no longer significant by month 16. More than half of those auto-renewed subscribers never visited the site again after the promotion, though. By month 12 the version that made people choose had more subscribers.</p>
<p>So a good share of readers saw the catch coming and stayed away. The authors suggest that a business which can credibly promise easy cancelling and timely reminders &quot;might end up with more consumers and larger revenues&quot;, and I think that&#39;s the line to take from it.</p>
<h3>When I&#39;d ask for the card</h3>
<p>When a buyer can see the product work within days and already means to pay if it does, such as an uptime monitor or a tool that books meetings, the card filters out the idly curious and the charge on day 8 lands on someone who wanted it.</p>
<p>I&#39;d skip it when the value takes weeks to show up, or needs setup and a pile of data before anything happens, or when most visitors arrive cold from search and have never heard of you. That&#39;s a lot of trust to ask for on a first visit. Whether you&#39;d lose 35% of starts, as the newspaper did, I honestly don&#39;t know; a software buyer isn&#39;t a newspaper reader, and nobody has published the test.</p>
<h2>How long a free trial should be</h2>
<p>Seven days is a sound default for software. Hema Yoganarasimhan, Ebrahim Barzegary and Abhishek Pani ran a field test with one software company across six markets, the UK and the US among them, giving 337,724 new users a 7, 14 or 30-day trial at random. The 7-day trial won 5.59% more subscriptions than the company&#39;s standard 30 days, and 14 days did no better than 30 (Management Science, 2023).</p>
<p>Why would less time sell more? Their data pointed at the end of the trial. People who went quiet for a long stretch near the end converted less, and a 30-day trial leaves plenty of room to go quiet.</p>
<p>Experienced users did gain from longer trials. Beginners did better with short ones.</p>
<p>The newspaper test leans faintly the other way. Four weeks against two lifted subscriptions by 4 to 5% during the promotion, though the authors found that gap wasn&#39;t statistically significant.</p>
<p>I&#39;d start at seven days and move only for a reason you can name, like a product whose first useful result needs a fortnight of data. Three days feels too tight once the reminder rules below apply, because the reminder has to land with time left to read it and act.</p>
<h2>What your site has to say before the trial starts</h2>
<p>Beside the button, before anyone commits, say what the trial turns into and when. For trials sold to consumers in the UK that&#39;s set to become law. The government&#39;s draft guidance, published in October 2026, calls a free or cheaper trial a &quot;concessionary period&quot; and lists the key information a buyer has to be given before signing up:</p>
<ul>
<li>that the plan renews and charges unless they end it before the trial is over, and on which day the first charge lands</li>
<li>how much that charge is, and how often it repeats</li>
<li>how to end the plan, and any notice needed</li>
<li>when you&#39;ll send reminders</li>
<li>their right to cancel in the cooling-off period</li>
</ul>
<p>Online, it has to be readable during sign-up without clicking a link, set apart from everything else and as close as possible to the moment they commit. And the last button has to make plain that pressing it means paying, or the buyer isn&#39;t bound.</p>
<p>Here&#39;s a made-up social scheduling tool at pagepost.example, the line under its button before and after:</p>
<table>
<thead>
<tr>
<th>Before</th>
<th>After</th>
</tr>
</thead>
<tbody><tr>
<td>Start free trial, then cancel any time</td>
<td>7 days free, then £12 a month from 24 November. We email you 3 days before. Cancel in Settings, any time</td>
</tr>
</tbody></table>
<p>The second one is three times as long and may leave whoever guards the page&#39;s whitespace feeling a bit henpecked. I&#39;d still ship it. The line under the button is where the last doubt gets answered, as it is for <a href="https://magpieaudit.com/articles/call-to-action-button-text">the words on the button itself</a>, and the price after the trial should match <a href="https://magpieaudit.com/articles/show-prices-on-landing-page">the price on your pricing table</a> to the penny, with yearly plans said as yearly and never dressed up as monthly.</p>
<h2>The reminder before the first charge</h2>
<p>Send it by email or another message the buyer can find again later. The UK guidance rules out pop-ups and push notifications that vanish, wants the purpose plain from the subject line, and wants the amount, the date, how to cancel and the date by which they need to act.</p>
<p>How far ahead is up to you, as long as your key information says it and the buyer gets &quot;a real opportunity&quot; to decide and cancel first.</p>
<p>After the first payment there&#39;s also a 14-day renewal cooling-off period. The buyer can still cancel in it, and may be owed a refund.</p>
<p>Which makes a charge from someone who forgot about the trial money you shouldn&#39;t count before it hatches, since they can cancel any time in the fortnight that follows and may get the money back (and I&#39;d expect plenty to, the moment they spot it on a statement).</p>
<p>Cancelling has to be straightforward, at any time, and online if signing up was. I&#39;d use a labelled button or a page in their account. An email address alone doesn&#39;t count, and the guidance doubts a compulsory &quot;why are you leaving&quot; form would either.</p>
<p>For what it&#39;s worth, Magpie&#39;s own plans start this way: seven days, the card at checkout, a reminder before the first charge, then 14 days to withdraw.</p>
<h2>Where the UK and US rules stand today</h2>
<p>Each of these was checked on 17 November 2026, and at least two of them could change within months. I&#39;m not a lawyer, and none of this is legal advice.</p>
<table>
<thead>
<tr>
<th>Rule</th>
<th>Status on 17 November 2026</th>
</tr>
</thead>
<tbody><tr>
<td>UK, DMCC Act 2024 subscription rules</td>
<td>Passed, not yet in force. Draft regulations laid on 18 October 2026, set to start on 1 January 2027 for contracts made from then on, still waiting for both Houses to approve</td>
</tr>
<tr>
<td>US, Restore Online Shoppers&#39; Confidence Act</td>
<td>In force since 2010</td>
</tr>
<tr>
<td>US, FTC click-to-cancel rule (2024)</td>
<td>Vacated by the Eighth Circuit on 8 July 2025; the old rule restored on 12 February 2026</td>
</tr>
<tr>
<td>US, a new FTC rule</td>
<td>Comments asked for in March 2026; no proposed rule published</td>
</tr>
</tbody></table>
<p>Once the UK rules apply, the Competition and Markets Authority can fine a business up to £300,000 or 10% of its global turnover, whichever is higher. A House of Lords committee drew the House&#39;s special attention to the regulations on 3 November 2026, so 1 January 2027 is the plan and not yet the law.</p>
<p>The US is messier. The 2024 click-to-cancel rule never took effect, because the court vacated it for &quot;procedural deficiencies&quot; in how the FTC made it rather than for anything it said, and the older rule put back in February 2026 covers only prenotification plans; the FTC&#39;s March notice says it doesn&#39;t cover &quot;continuity plans, automatic renewals, and trial conversions&quot;. So the law that bites is the Restore Online Shoppers&#39; Confidence Act. Since 2010 it has covered anything sold online with a charge that repeats unless the buyer acts. You have to disclose every material term before taking billing details, get express informed consent before charging, and give a simple way to stop the charges. The FTC also names California, Colorado, Vermont and Virginia among states with rules of their own.</p>
<p>If your site gives the UK&#39;s five points beside the button, I think you&#39;ve covered the main disclosures in both countries, though consent and the way out still live in your checkout and account pages, and a state you sell into may want its own wording on top.</p>
<h2>Questions people ask</h2>
<h3>Do the UK rules apply if I only sell to businesses?</h3>
<p>No, they cover subscriptions sold to consumers, not to businesses. If you sell to both, your sales to consumers are covered.</p>
<h3>Is a £1 trial treated differently from a free one?</h3>
<p>Not really. The guidance counts a cheaper trial as a concessionary period too, so you&#39;d still owe buyers the price after it, the date of the first full charge, the reminder and the cooling-off period. I&#39;d guess some teams pick a £1 trial thinking it gets them out of the rules, and it doesn&#39;t, and you&#39;ve added a charge to explain on top.</p>
<h3>Can the reminder be a notice inside the app?</h3>
<p>Not on its own. It has to be something the buyer can keep and read again, so an email or a text.</p>
]]></content:encoded></item>
<item><title>How fast a landing page should load on a phone</title><link>https://magpieaudit.com/articles/landing-page-speed-on-mobile</link><guid isPermaLink="true">https://magpieaudit.com/articles/landing-page-speed-on-mobile</guid><pubDate>Tue, 17 Nov 2026 21:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Search and AI</category><category>First screen</category><description>On a phone, the main picture or headline should show within 2.5 seconds, a tap should get a response within 200 milliseconds, and the layout should shift no more than 0.1, each for at least 75% of visits. Those are Google&apos;s Core Web Vitals. On a landing page, most of the gain comes from loading the main picture first, sizing every picture and moving other sites&apos; scripts later.</description><content:encoded><![CDATA[<p>Google publishes the number, which saves a lot of arguing. On a phone, the biggest thing on the first screen should be drawn within 2.5 seconds, a tap should get a visible response within 200 milliseconds, and nothing should jump about by more than a score of 0.1. Manage each of the three for at least three visits in four and your site passes.</p>
<p>My bet is that most of what turns PageSpeed Insights red on a landing page comes down to three or four causes. None of them needs a rebuild.</p>
<h2>The three numbers Google measures</h2>
<p>They&#39;re called Core Web Vitals, and each one times a different way a page can feel slow. These are web.dev&#39;s thresholds.</p>
<table>
<thead>
<tr>
<th>Measure</th>
<th>What it times</th>
<th>Good</th>
<th>Poor</th>
</tr>
</thead>
<tbody><tr>
<td>Largest Contentful Paint (LCP)</td>
<td>When the largest picture, video or block of text shows</td>
<td>2.5s or less</td>
<td>Over 4s</td>
</tr>
<tr>
<td>Interaction to Next Paint (INP)</td>
<td>How long the page takes to respond to a tap or a key</td>
<td>200ms or less</td>
<td>Over 500ms</td>
</tr>
<tr>
<td>Cumulative Layout Shift (CLS)</td>
<td>How far things move about while it loads</td>
<td>0.1 or less</td>
<td>Over 0.25</td>
</tr>
</tbody></table>
<p>On most landing pages the LCP is the hero picture, or the headline if there isn&#39;t one. A background video can be it too, since a <code>video</code> element counts and is timed by its poster image or its first frame, whichever shows first, so give it a small poster with its size set, or show a still on phones. INP watches taps, clicks and key presses for the whole visit and leaves scrolling and zooming out. It replaced First Input Delay in March 2024, so a guide still talking about FID is out of date.</p>
<p>CLS is the one you feel without knowing its name. You go to press &quot;Start free&quot;, a picture above it finishes loading (or <a href="https://magpieaudit.com/articles/cookie-banners-on-landing-pages">a cookie banner</a> slides in), and your thumb lands on a link instead. (A move within half a second of your own tap isn&#39;t counted, since you caused it.)</p>
<p>The pass mark is the 75th percentile, measured separately for phones and computers, and a page passes when each of the three is good for three visits in four. Which means the fast laptop on your office wifi tells you next to nothing about the slowest quarter of your phone visitors, and they&#39;re the people the number is about.</p>
<h2>Why your phone score is worse than your laptop&#39;s</h2>
<p>Phones are slower, and their connections shakier. The HTTP Archive&#39;s Web Almanac for 2025, built on July 2025 data, found 62% of mobile sites had a good LCP against 74% on desktop, and only 48% of mobile sites passed all three Core Web Vitals (up from 36% in 2023). Desktop sat at 56%.</p>
<p>So if you&#39;re red on mobile you&#39;re in the larger half. That&#39;s some comfort. Not much, though, if most of your visitors arrive on their phones from an ad.</p>
<h2>How to read a red PageSpeed Insights result</h2>
<p>The top of the PageSpeed Insights page and the bottom are measuring different things, and it&#39;s normal for them to disagree. The Core Web Vitals assessment at the top is field data, real Chrome visits to your site over the last 28 days from the Chrome UX Report. The score out of 100 underneath is lab data from one simulated load on a mid-tier Moto G4 over a mobile network.</p>
<p>Field data is what Google means by Core Web Vitals. The lab run is for finding causes, and its score wobbles between runs even when nothing&#39;s changed, so I&#39;d never chase it as a target. A 54 under a passing assessment is a site in decent shape. A 91 under a failing one isn&#39;t.</p>
<p>There are two gaps in the field data worth knowing about. A site needs enough visitors to be included at all, and Google doesn&#39;t say how many, so a brand new landing page often has no field data and only the lab run to go on (read it for its list of causes, then, and ignore the score until real visits turn up). And the Chrome UX Report only counts Chrome on desktop and Android. Chrome on iOS is left out entirely, so however many iPhone visitors you&#39;ve got, none of them are in those numbers.</p>
<h2>Does speed change how many visitors sign up?</h2>
<p>Probably, though the best evidence is weaker than the figures people quote from it. The study cited most is Deloitte&#39;s, commissioned by Google and published in 2020, which took 37 brands and 30 million sessions over four weeks, mostly in Europe and the US, and compared natural changes in speed with what visitors did. For retail sites, 0.1 seconds faster went with 8.4% more conversions. For travel it was 10.1%.</p>
<p>Read the lead generation part before those go in a deck. It covered six brands whose sites exist to get a form filled in, which is the closest thing in the study to most landing pages. Their mobile bounce rates fell 8.3%, and 21.6% more people got from the form&#39;s first step to sending it, yet their conversion rates on phones and computers both fell as the sites got faster. The report calls all of it correlation, and with 37 sites over four weeks I wouldn&#39;t read much more into it than that.</p>
<p>You&#39;ll also see &quot;53% of mobile visits are abandoned after 3 seconds&quot; pretty much everywhere. I couldn&#39;t find the original 2016 report, and Google&#39;s own pages credit the figure to two different studies, so I&#39;ve left it out.</p>
<p>For what it&#39;s worth, I think a slow first screen loses the visitors who were least sure in the first place, and they&#39;re exactly who <a href="https://magpieaudit.com/articles/landing-page-first-screen">a landing page&#39;s first screen</a> is there to win over. That&#39;s enough reason for me to fix the causes below without a percentage to point at.</p>
<h2>Does page speed affect Google ranking?</h2>
<p>A little, and relevance comes first. Google says &quot;Core Web Vitals are used by our ranking systems&quot; and, in the same documentation, that it &quot;always seeks to show the most relevant content, even if the page experience is sub-par&quot;. A good score &quot;doesn&#39;t guarantee&quot; a top place either. A fast page that doesn&#39;t answer the search will still lose to a slow one that does, which is why <a href="https://magpieaudit.com/articles/google-title-and-description">your Google title and description</a> usually deserve your time first.</p>
<h2>The fixes that matter most on a landing page</h2>
<p>A typical landing page is one long page with a big picture at the top, a few sections, a form or a button, and a handful of other companies&#39; scripts. That shape tends to fail in the same few places.</p>
<h3>Let the main picture load first</h3>
<p>The hero picture is usually your LCP, so anything that holds it up holds up the number. Web.dev&#39;s guide is unusually flat about this, saying &quot;Never lazy-load your LCP image.&quot; Lazy loading tells the browser to wait until it knows a picture&#39;s needed, which for the one at the top of the page is just delay, and the Web Almanac still found about 16% to 17% of pages doing it in 2025.</p>
<p>The fix is two attributes. Take <code>loading=&quot;lazy&quot;</code> off the hero image and add <code>fetchpriority=&quot;high&quot;</code>, which only about 17% of mobile pages with an LCP image used. One or two pictures at most, mind, because marking everything high is the same as marking nothing.</p>
<p>If the hero is a CSS background, the browser can&#39;t find it until the stylesheet has loaded. Preload it from the HTML with <code>&lt;link rel=&quot;preload&quot; as=&quot;image&quot; fetchpriority=&quot;high&quot;&gt;</code> and it&#39;ll start straight away.</p>
<h3>Give every picture its size</h3>
<p>A picture with no width and height takes up no room until it arrives, then lands like a gull on a chip and shoves everything below it down the screen. That&#39;s most of a bad CLS, and in 2025 62% of mobile pages had at least one picture with no size set.</p>
<p>Put <code>width</code> and <code>height</code> on every <code>img</code> and <code>video</code>, or give its box a CSS <code>aspect-ratio</code>. Anything that turns up late wants the same treatment (a cookie banner, a review widget, an embedded video), so keep its room with a <code>min-height</code> from the start rather than letting it push the headline down when it lands.</p>
<p>Fonts do it too. A web font that draws wider than the one shown before it reflows every line of the headline, and preloading the main font while matching its fallback&#39;s size with <code>size-adjust</code> keeps the words where they started.</p>
<h3>Move other sites&#39; scripts after the page</h3>
<p>Chat widgets, analytics, heatmaps, A/B testing tools and ad pixels often sit in the page&#39;s <code>&lt;head&gt;</code> as plain script tags. A plain script there stops the browser drawing anything until it&#39;s downloaded and run, so if one of those companies&#39; servers is slow that morning, your visitor stares at a white screen.</p>
<p>Add <code>defer</code> or <code>async</code> to them, or load them once the page has shown. The exception is a cookie consent tool, which has to run before the scripts it controls. Marketing might give you a beady look when their heatmap starts loading a second later. I&#39;d tell them anyway, and show them the white screen.</p>
<h3>Keep taps quick</h3>
<p>A slow INP on a landing page is usually those same scripts, running long jobs on the phone&#39;s single main thread just as the visitor presses the button, so the tap waits its turn. Fewer scripts is the first fix. In your own code, web.dev&#39;s advice is to break long work into pieces and hand control back to the browser often in between, and it goes as far as saying &quot;Yielding indiscriminately is better than not yielding at all&quot;.</p>
<h3>A worked example</h3>
<p>Here&#39;s fernly.example, a made-up plant care app, with the phone first screen these pages often have, before and after.</p>
<table>
<thead>
<tr>
<th>Before</th>
<th>After</th>
</tr>
</thead>
<tbody><tr>
<td>Hero photo with <code>loading=&quot;lazy&quot;</code> and no size</td>
<td>No lazy loading, <code>fetchpriority=&quot;high&quot;</code>, sized</td>
</tr>
<tr>
<td>Chat widget and heatmap scripts in the head, no <code>defer</code></td>
<td>Both deferred, the consent tool left as it was</td>
</tr>
<tr>
<td>Review widget appears above the button once it loads</td>
<td>Its height kept from the start</td>
</tr>
<tr>
<td>Headline font swaps in wider and reflows to three lines</td>
<td>Font preloaded, fallback matched with size-adjust</td>
</tr>
</tbody></table>
<p>None of that touches the design. Once loaded, the page looks exactly as it did before. It just gets there sooner and holds still, with <a href="https://magpieaudit.com/articles/call-to-action-button-text">its main button</a> staying put where the visitor&#39;s thumb expects it.</p>
<h2>What Magpie&#39;s free check measures</h2>
<p>Magpie&#39;s free check loads your site on an iPhone 14 Pro&#39;s screen and measures three of these causes. It records the layout shift while the phone&#39;s first screen loads, holds it to Google&#39;s 0.1 and names what moved. It finds the largest picture on that screen and checks it isn&#39;t lazy loaded and has its priority set (or, as a background, is preloaded). And it reads your HTML as the server sends it for other sites&#39; scripts in the head that hold the page up, leaving consent tools out. What it doesn&#39;t do is time a single load, since one visit from one place says very little about the slowest quarter of yours.</p>
<h2>Questions people ask</h2>
<h3>How long until a fix shows in my Core Web Vitals?</h3>
<p>Up to four weeks. The field data covers the last 28 days and updates daily, so a fix made today shows a little more each day until the old visits have rolled out of it.</p>
<h3>Is it worth paying for a faster host?</h3>
<p>Sometimes. Web.dev&#39;s guide to LCP splits it into four parts and, as a rough guide (it calls them &quot;guidelines, not strict rules&quot;), suggests about 40% of the time goes on the server&#39;s first answer, about 40% on downloading the picture and under 10% each on the two waits in between. So if your server is eating most of the 2.5 seconds by itself, a better host or a cache in front of it will help, and the lab run will show it as a slow first answer from the server. But on a landing page I&#39;d look at the hero picture and the scripts in the head first, because they cost nothing to fix and I&#39;d bet on them being the problem, and only then start pricing hosts.</p>
]]></content:encoded></item>
<item><title>Where visitors look first on a landing page</title><link>https://magpieaudit.com/articles/where-visitors-look-first</link><guid isPermaLink="true">https://magpieaudit.com/articles/where-visitors-look-first</guid><pubDate>Tue, 17 Nov 2026 21:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>First screen</category><category>Calls to action</category><description>Visitors look first at the top left of a screen and at text, and they scan plain unformatted text in an F shape. To get their eyes onto your headline and main button, make those two the largest, highest-contrast things on the first screen, keep them near the top left, and dim competing pictures and faces. A predicted heatmap guesses a typical visitor&apos;s first seconds; it doesn&apos;t record anyone.</description><content:encoded><![CDATA[<p>Top left, and words. That&#39;s where eyes went first across the 1,980 screens in the UEyes study (CHI 2023), whatever kind of screen it was, and text caught them more readily than pictures did.</p>
<p>Good news, then, if your headline already sits up there. Less good if your button is a pale outline over on the right next to a big smiling photo, because that&#39;s one of the likeliest things on the screen to be missed (and the fixes for it cost almost nothing).</p>
<h2>Where the eyes land in the first seconds</h2>
<p>UEyes sat 62 people in front of web pages, desktop apps, phone screens and posters for seven seconds each, and their eyes drifted high and left. The researchers found a top-left bias rather than the pull to the centre you might expect. Text drew more attention than images (they do point out their screens were mostly in English), and on web pages people liked to scan from left to right.</p>
<p>Nielsen Norman Group saw the same lean in ordinary browsing. Its 2017 study followed more than 120 people and over 130,000 fixations on wide monitors, and about 80% of those fixations fell on the left half of the screen, up from 69% in 2010.</p>
<p>Phones are stricter still. UEyes found people looked at fewer elements on a phone and went back to them less often, so whatever wins the first look there pretty much rules the roost. I&#39;d check your button on a phone before the laptop you built the page on, and if the page is slow to show up there at all, <a href="https://magpieaudit.com/articles/landing-page-speed-on-mobile">speed on mobile</a> comes first.</p>
<h2>What the F-pattern shows, and where it stops</h2>
<p>Jakob Nielsen described the F in 2006, from 232 users looking at thousands of pages. One sweep across the top, a shorter one a little lower down, then a run down the left side. It&#39;s how people scan a wall of plain text, and a landing page shouldn&#39;t hand them one.</p>
<p>Kara Pernice&#39;s 2017 update found it still held, on desktop and on mobile. She was specific about when it happens. The text has little or no formatting, the reader wants to be quick, and they aren&#39;t committed to reading every word. Nielsen Norman Group calls the result &quot;bad for users and businesses&quot;, because people may skip important content simply because it sits on the right.</p>
<p>Give people cues and they scan differently. Pernice lists the other patterns, the layer cake among them, where people read from heading to heading. For a first screen with a clear headline, one short line and one button, I&#39;d take one narrow habit from the F. Start every heading and line with the words that carry the meaning, because the left edge gets read and the end of a line often doesn&#39;t.</p>
<h2>Faces and pictures pull the eyes</h2>
<p>Faces win. In Cerf and colleagues&#39; 2007 study, people fixated a face with a probability of over 80% within their first two looks, which is about half a second. Those were photos of scenes rather than web pages, so I&#39;m guessing when I say a large face beside a headline would win the same race, though I&#39;d be surprised if it didn&#39;t.</p>
<p>Where the face is looking matters as well. Sajjacholapunt and Ball (2014) showed banner adverts to 72 people, and a face turned towards the advert&#39;s words raised attention to the whole banner and to its text, while a face staring straight out at the viewer kept the attention on itself. So if there&#39;s a person in your hero photo, turn them to the headline or the button and the picture takes the words under its wing.</p>
<p>Most other pictures get ignored. Nielsen&#39;s 2010 eye tracking found people skipped big feel-good images that were only there for decoration, and generic stock photos too, but they did study product photos and real people who work at the company. A picture of your product doing its job earns its space. A stock photo of a laptop on a desk just eats into the room for <a href="https://magpieaudit.com/articles/landing-page-first-screen">the first screen&#39;s words</a>.</p>
<h2>How to pull eyes onto the headline and the button</h2>
<p>Start with contrast. Itti and Koch&#39;s 2001 review of attention models says &quot;the perceptual saliency of stimuli critically depends on the surrounding context&quot;, which in plain terms means a button stands out by being different from what&#39;s around it. A black button on a white page pulls the eyes. Put the same black button on a dark photo and it&#39;s gone.</p>
<p>Here&#39;s a made-up first screen at ledgerly.example, a bookkeeping app, and what I&#39;d change on it.</p>
<table>
<thead>
<tr>
<th>On the screen</th>
<th>What it does to the eyes</th>
<th>What I&#39;d change</th>
</tr>
</thead>
<tbody><tr>
<td>A thin, light grey headline, centred</td>
<td>Low contrast, away from the top left</td>
<td>Dark, heavier and aligned left</td>
</tr>
<tr>
<td>A ghost button outlined in pale blue, &quot;Get started&quot;</td>
<td>Barely differs from the background</td>
<td>Solid fill in the brand&#39;s darkest colour</td>
</tr>
<tr>
<td>A large photo, right, a woman smiling at the camera</td>
<td>Takes the first looks and keeps them</td>
<td>Turn her towards the headline, or show the app</td>
</tr>
<tr>
<td>A blue promo strip above the header, &quot;50% off today&quot;</td>
<td>Looks like an advert, gets skipped</td>
<td>Move the offer into the line under the button</td>
</tr>
</tbody></table>
<p>The bottom row is what Nielsen Norman Group calls banner blindness. In Pernice&#39;s 2018 study anything boxed, animated or brightly treated like an advert got skipped, along with whatever sat next to it, so don&#39;t dress your one offer up as an ad.</p>
<p>Then size and place. The headline should be the biggest type on the screen, with the button tucked close under it so the two land in one glance rather than two. After that it&#39;s the button&#39;s words, since that&#39;s what a visitor reads once their eyes get there, and <a href="https://magpieaudit.com/articles/call-to-action-button-text">what your button should say</a> is a whole article of its own.</p>
<h2>Predicted heatmaps against real eye tracking</h2>
<p>A predicted heatmap is a model&#39;s guess at where an average person looks in their first few seconds with nothing in particular in mind. Real eye tracking records where actual people looked (a few dozen to a few hundred of them in the studies above, often given a task). Both are useful. Neither tells you who signs up.</p>
<p>Magpie&#39;s free report draws one of these predictions for each first screen, laptop and phone, over the first fifteen looks, about five seconds. It&#39;s worked out in code from a model we fitted on the UEyes recordings. On screens it wasn&#39;t fitted on, it matches where UEyes&#39;s viewers looked at a correlation of 0.50 on web pages and 0.59 on phone screens, where 1 would be perfect, and screen position alone gets 0.46 and 0.54, so place does a lot of the work and the rest is a modest gain. It doesn&#39;t detect faces yet.</p>
<p>It flags one thing in red, when the headline and the main button together hold under half of the strongest attention (the red on the heatmap). That threshold is our own judgement of where a first screen&#39;s attention ought to go. We haven&#39;t measured it against sign-ups, and I&#39;d rather say so plainly.</p>
<p>So treat a heatmap as a little bird&#39;s tip about a stranger&#39;s first glance, never a recording of your buyers. If the red sits on a photo, a promo strip or a loud row of <a href="https://magpieaudit.com/articles/customer-logos-on-landing-pages">customer logos</a>, pull it back with the changes above and check again. Then let your real visitors settle it with a test that counts sign-ups.</p>
<h2>Questions people ask</h2>
<h3>Does button colour matter for attention?</h3>
<p>Only through contrast, really. I&#39;d pick whichever colour differs most from what&#39;s directly around the button, then stop fiddling with it and spend the time on the words.</p>
<h3>Should the headline be centred or left-aligned?</h3>
<p>Honestly, either can work if it&#39;s large and dark. A short centred headline still sits near the top of the screen, which is most of what matters, and plenty of decent pages centre theirs because that&#39;s how the design falls. But both of the eye tracking studies above lean left, so if I were starting from a blank page I&#39;d put it on the left and only move it to the middle when the layout really asked for it, rather than because the template happened to come that way.</p>
]]></content:encoded></item>
<item><title>How many pricing plans should you offer?</title><link>https://magpieaudit.com/articles/how-many-pricing-plans</link><guid isPermaLink="true">https://magpieaudit.com/articles/how-many-pricing-plans</guid><pubDate>Tue, 17 Nov 2026 20:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Pricing</category><description>Most products do well with two to four plans, and three is a sound default, as long as each one suits a buyer you can name and differs from the next in a way that buyer can see. Mark the plan you&apos;d honestly recommend. Skip decoy plans built only to make another look cheap, and say plainly what yearly billing saves.</description><content:encoded><![CDATA[<p>Three is a fine place to start. Honestly, though, the number matters less than you&#39;d think, and I&#39;d take five well-told plans over three that blur into each other. What counts is that each plan belongs to a buyer you could describe in a sentence, and that a visitor can see within a few seconds what moving up a plan gets them.</p>
<p>If you&#39;re still deciding whether to put prices on your site at all, <a href="https://magpieaudit.com/articles/show-prices-on-landing-page">showing your prices</a> comes first. This one picks up once the plans are definitely going on your site.</p>
<h2>How many plans to offer</h2>
<p>Two to four covers most products. One works when every buyer needs the same thing. A fifth earns its place only if there&#39;s a fifth kind of buyer, say an agency running twenty clients where everyone else runs one.</p>
<p>I&#39;d write down who each plan is for before writing a single price. &quot;A café with one site and up to 15 staff&quot; is a buyer, whereas &quot;Pro&quot; is just a word. If two plans end up describing the same buyer, they&#39;re two eggs in one nest and one of them should go.</p>
<p>Here&#39;s a made-up shift rota tool for cafés at staffloop.example, worked through:</p>
<table>
<thead>
<tr>
<th>Plan</th>
<th>Who it is for</th>
<th>Price</th>
</tr>
</thead>
<tbody><tr>
<td>Solo</td>
<td>One café, up to 15 staff</td>
<td>£19 a month</td>
</tr>
<tr>
<td>Group</td>
<td>Up to 5 sites, shared staff</td>
<td>£59 a month</td>
</tr>
<tr>
<td>Chain</td>
<td>6 sites or more, payroll export</td>
<td>£149 a month</td>
</tr>
</tbody></table>
<p>A café owner can find themselves in that table without reading a single feature, because every row is a size of business they&#39;d recognise.</p>
<h3>What the famous jam study really found</h3>
<p>The jam study is a lot smaller than its fame. In 2000 Sheena Iyengar and Mark Lepper set up a tasting booth in a grocery store on two Saturdays, showing 24 jams for some hours and 6 for others. More people stopped at the big display, 145 of 242 passers-by against 104 of 260. Of those who stopped, though, nearly 30% bought jam after seeing six, against 3% after seeing 24. In actual people that&#39;s 31 buyers against 4.</p>
<p>Striking, and later work didn&#39;t bear it out as a rule. Benjamin Scheibehenne and colleagues pooled 50 experiments with 5,036 people in 2010 and found a mean effect of &quot;virtually zero&quot;, with wide swings between studies.</p>
<p>Then in 2015 Alexander Chernev and colleagues ran a second meta-analysis, over 99 observations and 7,202 people, and did find an effect once they sorted the studies by four things. Those were how complex the set of options was, how hard the task, how unsure people were of what they wanted, and whether they were trying to decide with as little effort as possible. It&#39;s the paper I&#39;d read if you&#39;re building a pricing table, because all four describe the person on your pricing page pretty much exactly. They don&#39;t know your product well yet. They&#39;re comparing columns of limits, and most of them would like to be done quickly. So I&#39;d rarely cut plans, and I&#39;d spend the time making them easy to tell apart instead.</p>
<h2>Why buyers pick the middle plan</h2>
<p>Itamar Simonson called it the compromise effect, back in 1989. In his studies a brand gained share when it became the middle option in a set, and more so when people expected to explain their choice to someone else. That last bit fits business software rather well, since a buyer who has to justify a purchase to their manager can always defend the middle plan.</p>
<p>Buyers tend to perch on the middle, so put the plan you most want to sell there, and make it the one you&#39;d honestly recommend to a typical customer. Then mark it. A badge like &quot;Most cafés pick this&quot; works much like a default, and defaults pull hard. Jon Jachimowicz and colleagues&#39; 2019 meta-analysis, across 58 studies with 73,675 people, put the effect size at 0.68 and found it stronger in consumer choices and where people read the default as a recommendation. Which means the badge has to be true. If it says most people pick it, most people should.</p>
<p>Simonson&#39;s studies had people choosing between described products in a study setting, and I couldn&#39;t find a published field test of three-plan pricing pages anywhere. I&#39;d treat the middle as a sound habit, then, and stop short of calling it a law.</p>
<h2>Do decoy plans work?</h2>
<p>Sometimes, in studies where every detail is a number, and that&#39;s about as far as the evidence goes. A decoy is a plan built to lose, worse than one of the others and barely cheaper, so that the other one looks like a bargain beside it. Joel Huber, John Payne and Christopher Puto first showed the effect in 1982, and in their studies the decoy itself was picked about 2% of the time.</p>
<p>Most pricing advice reaches for Dan Ariely&#39;s example instead. In his 2008 TED talk he describes an old offer from The Economist (online for $59, print for $125, or print and online for $125), which he gave to 100 MIT students before giving a version without the print-only option to another 100, and the most popular choice became the least popular. Those were students picking on paper with no money spent. A classroom demonstration, in other words, and a long way from a test on a live site.</p>
<p>Shane Frederick, Leonard Lee and Ernest Baskin looked harder in 2014. They couldn&#39;t replicate any of the five earlier studies they found that used realistic products, and across 19 studies of their own, where the options had to be experienced or looked at rather than read off as numbers, they found no effect. Shown as numbers, the same products brought it back. The original authors replied that same year that the effect holds when the original conditions are repeated, and fades when buyers already know what they want.</p>
<p>A pricing table is mostly numbers, so a decoy might shift a few choices on yours. I&#39;d still leave it out. Someone will buy the plan nobody should buy, by mistake, and then you&#39;ve got a customer who feels tricked writing in to support.</p>
<h2>How to lay out the table</h2>
<p>Nielsen Norman Group&#39;s 2024 guidance on comparison tables suggests them for up to five items, with column headers that stay on screen as you scroll (people &quot;will easily forget which column is for which product&quot;) and a way to hide the rows where every plan is the same.</p>
<p>The rows are where it tends to go wrong. Baymard Institute&#39;s 2021 study of software sites found plan tables listing &quot;on average 40 distinct features&quot;. Baymard&#39;s fixes are grouping features under labelled subheadings and putting the groups most buyers need first.</p>
<p>I&#39;d do five things:</p>
<ul>
<li>Put the plan names, prices and buttons in a header that stays visible as the table scrolls</li>
<li>Write each limit in words (&quot;up to 15 staff&quot;, &quot;5 sites&quot;), never a tick that leaves the buyer to guess the number</li>
<li>Lead with the two or three rows that really separate the plans, then group the rest</li>
<li>Order the plans from the smallest buyer to the largest, so a visitor finds their own size by reading along (I found no good test of left-to-right order, so this is only what reads naturally)</li>
<li>Give each plan&#39;s button words that say what happens next, as with any <a href="https://magpieaudit.com/articles/call-to-action-button-text">call to action button</a></li>
</ul>
<p>On a phone, five columns of ticks won&#39;t fit. Nielsen Norman Group suggests turning the columns into tabs there, so a visitor reads one plan at a time and switches between them.</p>
<h2>Monthly or yearly billing</h2>
<p>Whichever side a monthly and yearly switch opens on is a default, so everything above about defaults applies to it as well. It&#39;s worth having when a real share of your buyers would pay yearly. Open it on the billing you&#39;d recommend to most of them, and don&#39;t pick yearly just because it makes the monthly figure look smaller.</p>
<p>Say what yearly saves, in money. The UK&#39;s Competition and Markets Authority, in its November 2025 guidance on price transparency, says that where paying monthly and paying yearly cost different amounts, a trader may mislead buyers unless it makes that clear, for example by showing each option&#39;s total or by saying &quot;save £X&quot;. For a yearly plan paid up front it wants the full price in the offer itself. On staffloop.example that&#39;d be &quot;£190 a year, save £38&quot; beside &quot;£19 a month&quot;, and never a lone &quot;£15.83 a month&quot;. The guidance covers selling to consumers (and I&#39;m not a lawyer, so take this as a reading of it rather than legal advice), though I doubt a business buyer hit by a surprise £190 charge feels much different about it.</p>
<p>I&#39;m less sure than most pricing advice that a yearly plan keeps customers longer. The one field study I found comparing the two was at three US gyms, where Stefano DellaVigna and Ulrike Malmendier followed 7,752 members over three years, and members on a monthly contract were 17% more likely to stay enrolled beyond a year than those on a yearly one (which ran out unless renewed). The authors put it down to overconfidence, people expecting more of their future selves than they managed. A gym&#39;s a long way from software, and a yearly software plan usually renews by itself (as most <a href="https://magpieaudit.com/articles/free-trial-card-up-front">free trials that take a card</a> do), so it proves little either way. I&#39;d read it as a reminder that a yearly discount buys you cash up front, and not necessarily loyalty.</p>
<h2>What a pricing check looks at</h2>
<p>Magpie&#39;s full report reads a site&#39;s plans beside what its rivals charge, found by web search. It looks for limits stated in words, a marked usual choice and buttons that say what each plan starts, and it compares like with like, so a per-seat price is set against a flat one at one stated team size, and a plan billed monthly or yearly counts as one plan.</p>
<h2>Questions people ask</h2>
<h3>Is an Enterprise column worth adding?</h3>
<p>Only when large buyers really do need a different contract, their own security review, say, or invoicing terms your other plans don&#39;t offer. If you add one, I&#39;d still give it a starting price or a typical one (for the reasons in <a href="https://magpieaudit.com/articles/show-prices-on-landing-page">showing your prices</a>), and I&#39;d keep it a bit out of the way, maybe as a line under the table rather than a fourth column, because the people choosing between your main plans are most of your buyers and the Enterprise column mostly just sits there taking up a quarter of the width. That&#39;s a judgement call, mind. If big contracts are where most of your money comes from, give it the full column.</p>
<h3>How big should the yearly discount be?</h3>
<p>I found no research figure to copy. The gyms in that study gave two months free, about 17% off a year of monthly fees, which is a common enough offer. Whatever you choose, say the saving in money and make sure it&#39;s true.</p>
]]></content:encoded></item>
<item><title>Star ratings and review badges on a landing page</title><link>https://magpieaudit.com/articles/ratings-and-review-badges</link><guid isPermaLink="true">https://magpieaudit.com/articles/ratings-and-review-badges</guid><pubDate>Tue, 17 Nov 2026 19:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Social proof</category><description>Show a star rating, review count or award badge only when a visitor can check it: name the site it comes from, link to its page there, keep the figure as that page shows it today, and retire a badge once its season is long over. A true 4.6 can sell better than a polished 5.0, and in the UK a misleading rating is banned outright.</description><content:encoded><![CDATA[<p>Five stars and a number fit on one line under a headline, and a visitor reads them in a second. That&#39;s why they get copied from page to page, rounded up a little, and left in place long after the figure behind them has moved on.</p>
<p>On 11 November 2026 we read 87 of the landing pages we&#39;d been testing our reports on. It&#39;s a hand-picked set rather than a fair sample of the web, so take the share loosely, but about 37 of them showed at least one rating, review count, award badge or rating hidden in the page&#39;s code for search engines. Among the ones that wouldn&#39;t survive a check were stars that named no source and ratings kept in the code where no visitor could see them. Both are quick fixes.</p>
<h2>Why a 4.6 can sell better than a 5.0</h2>
<p>Researchers at Northwestern University&#39;s Spiegel centre found in 2017 that purchase likelihood typically peaks for ratings between 4.0 and 4.7, &quot;and then begins to decrease as ratings approach 5.0&quot;, across product categories. The same report found a product with five reviews was 270% more likely to be bought than one with none.</p>
<p>That was shop products, not software or services, and the study doesn&#39;t say why. My guess is that a buyer who sees 5.0 starts wondering who left the bad reviews out. So I&#39;d show the real 4.6, with its count and its source, before I&#39;d show a round 5.0 with nothing beside it, because a careful buyer is going to wonder about a flawless score regardless.</p>
<h2>Name the site every figure comes from</h2>
<p>&quot;★★★★★ Loved by 1,000+ teams&quot; says nothing about who did the loving, or where. A visitor has to take it on trust, and a landing page is about the last place they&#39;ll do that.</p>
<p>UK law takes these figures as seriously as the reviews themselves. The Competition and Markets Authority&#39;s guidance on <a href="https://www.legislation.gov.uk/ukpga/2024/13/schedule/20">the Digital Markets, Competition and Consumers Act</a> counts &quot;overall ratings, overall summaries, review counts and rankings&quot; as consumer review information, and the Act bans publishing it &quot;in a misleading way&quot;. A business that shows on its own website the rating it got on a review site is, in the guidance&#39;s words, a &quot;second-party publisher&quot; with a duty of its own, one it can&#39;t hand over to that site. I&#39;m not a lawyer, and this isn&#39;t legal advice, but the guidance is plain enough on that point.</p>
<p>The review sites want the source named as well. Trustpilot&#39;s brand guidelines ask for the word &quot;TrustScore&quot; beside the number, and its logo with the rating. G2 wants a badge&#39;s &quot;full badge name (including segment or region, if applicable)&quot;.</p>
<p>Here&#39;s a made-up booking app at tablefor.example, before and after:</p>
<table>
<thead>
<tr>
<th>Before</th>
<th>After</th>
</tr>
</thead>
<tbody><tr>
<td>★★★★★ Rated 5 stars by restaurants</td>
<td>TrustScore 4.6 from 312 reviews on Trustpilot, linked to its profile</td>
</tr>
<tr>
<td>G2 Leader</td>
<td>G2 Leader, Restaurant Management, Small Business, Summer 2026, linked to its category</td>
</tr>
<tr>
<td>1,000+ five-star reviews</td>
<td>4.7 from 1,180 ratings on the App Store, linked to its listing</td>
</tr>
</tbody></table>
<h2>Link it to the page where it lives</h2>
<p>Baymard Institute&#39;s 2023 testing of tours and experiences sites watched people click third-party rating badges, expecting the reviews behind them. Where the badge went nowhere, some left the site to search for the reviews on the rating site themselves. One called it a &quot;dead link&quot;. Another said &quot;this is not a link. This is just like a screenshot.&quot;</p>
<p>Trustpilot asks for the link in so many words (&quot;Link back to your Trustpilot profile page when using our brand on your website&quot;), and G2 wants a backlink to its category page for every badge. I think it&#39;s the same doubt that makes visitors trust quotes on outside sites more than <a href="https://magpieaudit.com/articles/landing-page-testimonials">testimonials on your own site</a>, and a link is about the cheapest way there is to answer it on your own page.</p>
<h2>Keep the figure as it stands today</h2>
<p>On 14 October 2026 the Advertising Standards Authority ruled against a jewellery retailer whose product page showed &quot;4.7/5 | 2540+ reviews&quot; beside Trustpilot&#39;s name. When the ASA looked on 4 August 2026, the Trustpilot profile had &quot;a rating of 4.1 from 1,662 reviews&quot;, and the retailer accepted it hadn&#39;t updated the claim.</p>
<p>The CMA&#39;s guidance says something similar about fake reviews, that a star rating or count not &quot;promptly&quot; updated once fakes are removed is published in a misleading way. Any figure typed into your site&#39;s code stops being true the day the real one moves. You can use the review site&#39;s own live widget, or check the number against the source by hand once a month. The second is duller, but it works on every page builder going.</p>
<h2>Careful with &quot;#1 rated&quot;</h2>
<p>In October 2025 the ASA ruled against an internet provider that called itself the &quot;#1 rated internet provider on Trustpilot&quot;. Its own rating was 4.6, and the top provider in that Trustpilot category had 4.9. It only came first after filtering to London and 4.5 stars or more, then sorting by the number of reviews, which took about 400 companies down to 20.</p>
<p>The ASA read the claim the way a visitor would, as highest rated against all its rivals. If yours needs a footnote explaining the filters, it&#39;s a smaller claim than it looks, and I&#39;d just make the smaller claim (the same goes for <a href="https://magpieaudit.com/articles/comparing-yourself-with-rivals">&quot;cheapest&quot; and &quot;the only&quot;</a>, which read as claims about every rival at once).</p>
<h2>Award badges have a season</h2>
<p>G2&#39;s guidelines recommend that references &quot;Reflect current quarter and year&quot;, and its real-time rankings should carry &quot;the date of when the rankings were pulled&quot;. So a &quot;Leader Winter 2023&quot; badge in late 2026 makes a careful buyer wonder whether the product has placed since, which isn&#39;t much to strut like a peacock over.</p>
<p>Launch awards are a bit different. &quot;Product of the Day&quot; names one day and stays true about that day, so it can stay as long as it says which launch site gave it and links there. I&#39;d still move it below the fold once newer proof exists, since it&#39;s a fledgling&#39;s prize from launch week and says little about the product now.</p>
<h2>The rating in your code counts too</h2>
<p>Plenty of sites carry a rating in their structured data, the hidden markup that asks Google for stars in <a href="https://magpieaudit.com/articles/google-title-and-description">its search results</a>. Google&#39;s rules for it are short. &quot;If you use AggregateRating, users should be able to see that aggregate rating on the page&quot;, and &quot;Don&#39;t aggregate reviews or ratings from other websites.&quot; Since September 2019 it has also shown no stars for a business or organisation marked up as reviewing itself on its own site (&quot;a review about entity A is placed on the website of entity A&quot;).</p>
<p>Of those 87 pages, 15 carried this markup, and some held figures such as 5.0 from 86,000 ratings that no visitor could see anywhere on the page. I can&#39;t see the upside in that. Show the figure with its source, or take the markup out.</p>
<h2>Show the good with the bad</h2>
<p>Picking only the five-star reviews makes your rating harder to believe, and in the US and UK it can break the rules. The FTC&#39;s rule bars a business from implying the reviews it displays are most or all of those submitted while it suppresses some for their low rating or negative words (16 CFR 465.7). The UK Act names &quot;failing to publish, or removing from publication, negative consumer reviews whilst publishing positive ones&quot; as one way of publishing misleadingly, and Trustpilot&#39;s own list of don&#39;ts includes &quot;displaying only five star reviews without a disclaimer&quot;.</p>
<p>You can still lead with your best ones. Put the overall rating beside them, linked to the rest, so a visitor can see you&#39;re not hiding the 3-star reviews.</p>
<h2>What Magpie checks</h2>
<p>Magpie&#39;s full report reads every rating, review count and award badge on your site by code, with no model involved. It checks that each names the site it comes from, links to it and is current (a seasonal badge whose season ended over a year ago gets flagged). An App Store rating is compared with Apple&#39;s own figure for the app, and a rating in the page&#39;s code that no visitor sees is flagged too. Review sites that turn automated visits away, as the big ones do, are taken as shown and never got round.</p>
<h2>Questions people ask</h2>
<h3>Do I need a paid plan to show a Trustpilot or G2 badge?</h3>
<p>Usually, yes. Trustpilot&#39;s guidelines say businesses not on a Trustpilot plan &quot;aren&#39;t allowed to display their TrustScore or Star Rating on their websites&quot;. G2&#39;s free profiles can&#39;t show badges from its Summer 2025 reports onwards, though they keep earlier ones and the &quot;Users Love Us&quot; badge.</p>
<h3>Should I show a rating under 4?</h3>
<p>I wouldn&#39;t put it on the first screen. Spiegel&#39;s figures put the peak between 4.0 and 4.7, so below that the number is probably working against you, though I&#39;d add that it was shop products they looked at, and a 3.8 for a niche bit of software with twelve reviewers might read quite differently to someone who knows how few people review that sort of thing. Either way I&#39;d lead with two or three specific, named quotes instead, and keep the rating lower down for anyone who goes looking.</p>
<h3>Where should the rating go on my site?</h3>
<p>Near the claim or the button it backs up. On <a href="https://magpieaudit.com/articles/landing-page-first-screen">the first screen</a>, one short line with its source beside <a href="https://magpieaudit.com/articles/call-to-action-button-text">the main button</a> is plenty, and a wall of badges can sit further down for whoever&#39;s still weighing it up.</p>
]]></content:encoded></item>
<item><title>Should you show prices on your landing page?</title><link>https://magpieaudit.com/articles/show-prices-on-landing-page</link><guid isPermaLink="true">https://magpieaudit.com/articles/show-prices-on-landing-page</guid><pubDate>Tue, 17 Nov 2026 19:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Pricing</category><category>First screen</category><description>Yes. Show your prices on your site, or at least a starting price for each plan, because buyers who can&apos;t find one assume the worst and check a rival who shows theirs. Where a price really is set per deal, show what a typical customer pays. Keep &quot;Contact sales&quot; for the largest contracts, never for every plan.</description><content:encoded><![CDATA[<p>Show them. Someone who can&#39;t find your price won&#39;t sit tight for a sales call to hear it. They&#39;ll guess, then open a rival&#39;s pricing page to see whether the guess was fair, and you may not get them back. What follows is how to show prices well (plans, yearly billing, fees) and the fairly narrow case where &quot;Contact sales&quot; is the honest answer.</p>
<h2>Why a hidden price sends buyers to a rival</h2>
<p>Hoa Loranger at Nielsen Norman Group was blunt about it in 2013. Drawing on numerous usability studies with business customers, she wrote that they report pricing as &quot;the top most needed piece of information online&quot;, and that when a site left it out, &quot;participants go to competitors&#39; sites&quot;. It costs more than the one visit, too. In her words, &quot;People view companies that hide costs as being evasive and untrustworthy.&quot; A hidden price can make a perfectly honest business look a bit fly-by-night.</p>
<p>Therese Fessenden&#39;s 2019 article for the same group has the bit I&#39;d worry about most. The first site to show a price can anchor what buyers expect to pay, so if a rival publishes theirs and you don&#39;t, theirs is likely to be the number your buyer has in mind when they reach you.</p>
<p>Shoppers say much the same at a checkout. In Baymard Institute&#39;s research, 12% of US online shoppers who abandoned an order said they &quot;couldn&#39;t see / calculate total order cost up-front&quot;, and 40% left because extra costs were too high (the commonest reason, once &quot;just browsing&quot; is set aside). That&#39;s retail. I wouldn&#39;t lean on the exact numbers for a software landing page, but a price found late, or never, losing people who were ready to buy seems to me to carry over pretty well.</p>
<h2>When &quot;Contact sales&quot; is fair</h2>
<p>For the part of your pricing that&#39;s genuinely set deal by deal, and nothing else. A 2,000-seat contract with its own security review and support terms gets priced in a meeting, and a buyer at that size expects one. A plan for a team of five doesn&#39;t. A small buyer takes to a clear price like a duck to water, and putting it behind a form adds a step between someone and a decision they could have made alone, on a Sunday evening, without ever speaking to you.</p>
<p>Even the custom part can carry a number. Jakob Nielsen&#39;s advice from 2006 still holds. Where the exact price depends on the customer, he wrote, &quot;offer users some representative cases and their prices&quot;. His example was overnight shipping. One carrier wanted ZIP codes, a weight and box sizes before it would quote anything at all, while a rival simply showed a small table of prices for common parcels, which was all most people needed, since those drawing up a shortlist &quot;needed only a general idea of cost levels&quot;.</p>
<p>Here&#39;s a made-up invoicing tool for agencies at ledgerly.example, before and after:</p>
<table>
<thead>
<tr>
<th>Plan</th>
<th>Before</th>
<th>After</th>
</tr>
</thead>
<tbody><tr>
<td>Solo</td>
<td>£12 a month</td>
<td>£12 a month, 1 user</td>
</tr>
<tr>
<td>Team</td>
<td>Contact us</td>
<td>£39 a month, up to 10 users</td>
</tr>
<tr>
<td>Enterprise</td>
<td>Contact us</td>
<td>From £900 a month for 50 users; a typical 120-user agency pays about £1,900</td>
</tr>
</tbody></table>
<p>The biggest buyers still end up on a call. They just turn up already knowing roughly what it&#39;ll cost them.</p>
<h2>How to lay out the plans</h2>
<h3>Mark the plan most buyers pick</h3>
<p>Jon Jachimowicz and colleagues pooled 58 studies of defaults in 2019, with 73,675 people between them, and found defaults had &quot;a considerable influence&quot; on what people chose, more in consumer choices than in environmental ones. Defaults were stronger when people read them as a recommendation. I think a &quot;Most teams pick this&quot; badge is about as close to a spoken recommendation as a pricing table gets.</p>
<p>So mark the plan you&#39;d honestly point most buyers towards, and give each plan a few words on who it suits. A badge on the dearest plan, with no reason a buyer can see, just steers people towards what suits you, and I&#39;d take it off.</p>
<h3>Three plans or five matters less than how they differ</h3>
<p>In 2010 Benjamin Scheibehenne and colleagues pooled 50 experiments with 5,036 people and found that, on average, more options had close to no effect on choosing, though the results swung a lot from one study to the next. My guess is that options which are hard to compare do more harm than the count of them. I&#39;ve gone into <a href="https://magpieaudit.com/articles/how-many-pricing-plans">how many pricing plans</a> to offer separately, and the grid of ticks comes up there as well.</p>
<p>Write each limit in words (&quot;up to 10 users&quot;, &quot;5 client portals&quot;, &quot;invoices in 3 currencies&quot;) so someone can see what moving up a plan buys without reading every row.</p>
<h3>Say what yearly billing costs</h3>
<p>Picture &quot;£32 a month&quot; in large type with &quot;billed yearly&quot; in small grey type under it. Anyone skimming reads that as a monthly plan, and the first charge of £384 lands as a nasty surprise, worse still at the end of <a href="https://magpieaudit.com/articles/free-trial-card-up-front">a free trial that took a card</a>. In my view it&#39;s the commonest way an honest price ends up unclear.</p>
<p>The UK&#39;s Competition and Markets Authority says in its November 2025 guidance on price transparency that a business offering monthly and yearly payment for the same product risks misleading buyers unless the difference in cost is clear, for example by giving the total of each option or a line such as &quot;save £X&quot;. So show both plainly, &quot;£32 a month, billed £384 a year&quot; beside &quot;£40 billed monthly&quot;, and let the buyer do the sum if they want to.</p>
<h3>Put every fee in the number</h3>
<p>If you sell to consumers in the UK, the price you show (and any <a href="https://magpieaudit.com/articles/site-promises-and-your-terms">&quot;no hidden fees&quot; promise</a> beside it) has to include every charge they can&#39;t avoid, such as VAT, a set-up fee or a platform fee. The same CMA guidance calls leaving one out of the first price and adding it later drip pricing, and the law prohibits it. For a monthly plan with a one-off set-up fee, it asks for a first month&#39;s price that includes the fee, plus a line saying what later months cost. Even &quot;from £12&quot; counts, as the guidance treats a lowest price as an invitation to buy.</p>
<p>The guidance is about selling to consumers, so a site that sells only to other businesses sits outside it, though I&#39;d still write &quot;excluding VAT&quot; right beside any price that leaves it out. A fee that only turns up at checkout still loses the order, mind, which is more or less what Baymard&#39;s 40% were telling it. (I&#39;m not a lawyer, and none of this is legal advice.)</p>
<h2>Where the price sits on a landing page</h2>
<p>The full plans table goes further down, once a visitor knows what the product does, since the first screen has three other jobs to do before prices come into it. What does earn a place up top is one short line. Something like &quot;Free for 3 clients, then £12 a month&quot; under the main button answers the doubt a visitor has at the moment of clicking, which is the same job <a href="https://magpieaudit.com/articles/call-to-action-button-text">the line under the button</a> does for a card or a time. Then add a &quot;Pricing&quot; link to the header so the table&#39;s one press away from anywhere.</p>
<p>Beside the plans, I&#39;d put proof about the worries people have after paying, such as a named customer&#39;s quote about the support they got or how easy it was to cancel.</p>
<p>Apps sold through the App Store are the exception. Apple works out prices for the other 174 storefronts, in 43 currencies, from one base price and adjusts them now and then for taxes and exchange rates, so any single price on your website won&#39;t match what many readers see in their own store. Say what&#39;s free and what the paid upgrade adds, and treat a &quot;from&quot; price as a bonus if you want one.</p>
<h2>What a pricing check looks at</h2>
<p>Magpie&#39;s full report finds what a site&#39;s rivals charge by web search and reads its pricing beside theirs, comparing like with like (a per-seat price against a flat one at a stated team size, a yearly plan as yearly). It also reads the pricing page and terms against what the UK&#39;s subscription rules (drafted to start on 1 January 2027) say a buyer must be told before paying, which is that it renews, how much and how often, and how to end it.</p>
<h2>Questions people ask</h2>
<h3>What if my prices are higher than my rivals&#39;?</h3>
<p>Show them anyway, with what the extra buys written right next to the number. People <a href="https://magpieaudit.com/articles/comparing-yourself-with-rivals">compare you with rivals</a> whether your price is on the page or not. Hide it and they&#39;re comparing the rival&#39;s real price with their own guess at yours, and I wouldn&#39;t bet on that guess going your way.</p>
<h3>Is a pricing calculator a good idea?</h3>
<p>As an extra, maybe, though I&#39;d never make it the only way to get a number. Loranger found the pricing tools in her studies &quot;complex, time-consuming, and error prone&quot;, and it was mostly customers who were already keen who bothered with them, which makes sense when you think about who&#39;s on your pricing page. Someone with five rivals open in other tabs isn&#39;t going to sit there entering seat counts and usage guesses just to learn if you&#39;re anywhere near their budget. A small table of common cases, like the ledgerly.example one above, answers more people faster. Keep the calculator underneath it for the ones who want the exact figure.</p>
]]></content:encoded></item>
<item><title>What your Google title and description should say</title><link>https://magpieaudit.com/articles/google-title-and-description</link><guid isPermaLink="true">https://magpieaudit.com/articles/google-title-and-description</guid><pubDate>Tue, 17 Nov 2026 17:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Search and AI</category><category>First screen</category><description>A landing page&apos;s title should say what the product does in the buyer&apos;s words, then the site&apos;s name, short enough to fit a result&apos;s width, since Google cuts by width rather than characters. The meta description should sum up the page in one or two sentences. Google rewrites titles that are half-empty, out of date, inaccurate or repeated across pages.</description><content:encoded><![CDATA[<p>Your result in Google is a name, a blue line and a couple of grey ones. The blue line is your title, or whatever Google wrote in its place. The grey lines are your meta description, or a bit of your page that Google picked instead. If either comes out cut off mid-word or rewritten, the cause is usually something on the page you can change.</p>
<p>Getting <a href="https://magpieaudit.com/articles/named-in-ai-answers">named in AI answers</a> overlaps with this, but it brings crawlers and rules of its own. This one&#39;s about the blue link.</p>
<h2>Google cuts by width, not characters</h2>
<p>Google doesn&#39;t set a limit on either. Its documentation says they&#39;re shortened &quot;as needed, typically to fit the device width&quot;, and it doesn&#39;t publish what that width is in pixels.</p>
<p>Which is why counting characters misleads people. In Arial at 20px, roughly a desktop title&#39;s size, a capital W takes almost 19 pixels and a lower case i under 5, so a title made of Ws would run out of room after about 30 characters and one made of i&#39;s would last about 130. Nobody writes either, obviously. Real titles land somewhere in between, and it tends to be one long word near the end that tips them over.</p>
<p>The figure most tools use, and the one I&#39;ve used in the example further down, is about 580 pixels for a desktop title and about 920 for the description. It&#39;s an estimate and I&#39;d treat it as one. If what matters sits in the first half of the title, a cut won&#39;t cost you much.</p>
<h2>When Google rewrites your title</h2>
<p>In August 2021 Google changed how it writes title links, and its Search Liaison, Danny Sullivan, said the <code>&lt;title&gt;</code> element was still the source &quot;more than 80% of the time&quot;. After feedback, a follow-up post in September put it at about 87%. Those are 2021 figures and neither post gives a newer one, so I&#39;d guess the share has moved since, without being able to say which way. The same system &quot;generally&quot; stopped changing titles by the query, though snippets still change with every search.</p>
<p>The September post named four cases where Google goes looking for other words.</p>
<table>
<thead>
<tr>
<th>Your title is</th>
<th>What Google does</th>
</tr>
</thead>
<tbody><tr>
<td>Half-empty, &quot;| Site Name&quot;</td>
<td>Fills in the product&#39;s name from the page</td>
</tr>
<tr>
<td>Out of date, &quot;2020 ...&quot;</td>
<td>Puts in the year the page&#39;s own headline shows</td>
</tr>
<tr>
<td>Inaccurate</td>
<td>Cuts it to what the page actually holds</td>
</tr>
<tr>
<td>The same on several pages</td>
<td>Adds what tells each page apart, such as &quot;Season 1&quot;</td>
</tr>
</tbody></table>
<p>Today&#39;s documentation adds more. One is a page with several big headings and no clear main one. Another is a title in a different language, or a different alphabet, from the page itself. A third is the site&#39;s name repeated. And a title with last year&#39;s date in it is a lame duck already.</p>
<p>When it does rewrite, Google draws on the main title a visitor sees, the <code>&lt;h1&gt;</code> and other headings, <a href="https://magpieaudit.com/articles/shared-link-previews"><code>og:title</code></a>, large prominent text, the words of links pointing to the page and <code>WebSite</code> structured data. I haven&#39;t found a switch to turn rewriting off. You can give it no reason to, though.</p>
<h2>Put your site&#39;s name last, and briefly</h2>
<p>Google&#39;s advice is to &quot;brand your titles concisely&quot;, with the site&#39;s name once, at the start or the end, after a hyphen, colon or pipe. A longer brand line repeated on every page, it warns, &quot;will look repetitive&quot;.</p>
<p>On a landing page I&#39;d put the name last, and I don&#39;t think it&#39;s a close call. The words before it are what a stranger reads first, and the name already shows above the title as the result&#39;s site name anyway. Google even says it may drop a repeated name from the title link for that exact reason, where the site name is set for the whole domain.</p>
<p>That site name above the title comes from your home page, and Google picks it by itself. You can say which name you&#39;d prefer with <code>WebSite</code> structured data on the home page, giving its <code>name</code> and <code>url</code>. It works for a domain or a subdomain but never for a folder such as <code>/product</code>, which is worth knowing if your product lives under a bigger company&#39;s site.</p>
<h2>Make the title and the headline say the same thing</h2>
<p>Google&#39;s sources for a replacement title include the main title shown on the page and its <code>&lt;h1&gt;</code>, and an inaccurate title is one of its named reasons to swap yours out. So if your title and your headline disagree, you&#39;ve handed Google a reason and a ready-made alternative in one go.</p>
<p>I&#39;d write <a href="https://magpieaudit.com/articles/landing-page-first-screen">the first screen&#39;s headline</a> first, about what the product does for the visitor, then let the title carry the same promise in fewer words. They don&#39;t need to match word for word (a headline can afford a few more words than a title) but a visitor should be able to tell they&#39;re about the same thing.</p>
<h2>What the meta description should say</h2>
<p>Most snippets are built from the page itself. Google uses your meta description when it&#39;s &quot;a more accurate description of the page than content taken directly from the page&quot;, and the snippet can change from one search to the next, as Google shows whichever part fits the query best.</p>
<p>So your description is competing with your own page for the job. The SEO Starter Guide asks for &quot;a succinct, one- or two-sentence summary of the page&quot;, unique to it and holding its most relevant points. The bad example in Google&#39;s documentation is a sewing shop&#39;s string of keywords, from &quot;Sewing supplies, yarn&quot; on through machines, threads, bobbins and needles, and Google says a description like that is less likely to be shown at all.</p>
<p>Put the outcome first and then the detail that settles a doubt, such as a price (the same case I&#39;d make for <a href="https://magpieaudit.com/articles/show-prices-on-landing-page">showing prices on the landing page itself</a>) or &quot;free for up to five plants&quot;. If the second sentence gets cut, the first still stands on its own.</p>
<h2>A worked example</h2>
<p>Here&#39;s fernly.example, a made-up plant care app, measured the way a desktop result would draw it.</p>
<table>
<thead>
<tr>
<th>Part</th>
<th>Words</th>
<th>Width</th>
</tr>
</thead>
<tbody><tr>
<td>Title, before</td>
<td>Fernly | The future of plant care | Plant watering app, plant reminders, plant care</td>
<td>709px, cut</td>
</tr>
<tr>
<td>Title, after</td>
<td>Know when each plant needs water | Fernly</td>
<td>388px</td>
</tr>
<tr>
<td>Description, before</td>
<td>Plant care app, watering reminders, plant app, house plants, plant schedule, plant tracker, best plant app, plant reminder app, garden app, indoor plants</td>
<td>939px, cut</td>
</tr>
<tr>
<td>Description, after</td>
<td>Snap a photo of each plant and get a reminder on your phone the day it needs watering. Free for up to five plants.</td>
<td>704px</td>
</tr>
</tbody></table>
<p>The old title gets cut after &quot;plant&quot;, spends its first words on a slogan and says &quot;plant&quot; four times, like a woodpecker stuck on one tree. The new one is just the page&#39;s headline with the name after it, and it&#39;s got room to spare. The old description is the sewing shop&#39;s mistake again in a different trade, while the new one would still make sense if Google lost its second sentence.</p>
<h2>Checking what your result will say</h2>
<p>Open the page&#39;s source and you can read both, then measure them against those widths. Read it as the server sends it, mind, and not the page after its scripts have run, because a title a script sets later is easy to miss.</p>
<p>Magpie&#39;s full report does this in code, in its &quot;In search results&quot; section. It reads the title and description from the HTML the server sends, measures each in Arial&#39;s pixels and says where Google would cut it, or that the description is missing or too short to say much.</p>
<h2>Questions people ask</h2>
<h3>Should every page have its own description?</h3>
<p>Yes, and Google&#39;s fairly direct about it, saying identical or similar descriptions on every page &quot;aren&#39;t helpful&quot;. It suggests a description of the whole site on the home page and pages that gather others, and one of its own on every other page. My guess is that most repeated descriptions come from a template that fills the field in once for the whole site, so the pricing page, the blog and the landing page all end up saying the same sentence, and nobody spots it because nobody reads the head of their own pages. It&#39;s a five-minute check, and a slightly embarrassing one if the same line turns out to have sat on every page for years.</p>
<h3>Can I stop Google showing a snippet?</h3>
<p>You can. A <code>nosnippet</code> robots meta tag stops it, <code>max-snippet</code> sets its longest length and <code>data-nosnippet</code> keeps one part of the page out. On a landing page I&#39;d rarely bother, since each one takes away words a searcher could&#39;ve read.</p>
]]></content:encoded></item>
<item><title>What makes a landing page testimonial believable</title><link>https://magpieaudit.com/articles/landing-page-testimonials</link><guid isPermaLink="true">https://magpieaudit.com/articles/landing-page-testimonials</guid><pubDate>Tue, 17 Nov 2026 17:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Social proof</category><category>First screen</category><description>A testimonial&apos;s believable when a real person a visitor could look up says something specific. Give each quote the person&apos;s name, their role or company, a real photo and a link to where they said it, and keep its words to a concrete use or result. Vague praise from nobody in particular is easy to fake, visitors discount it, and invented quotes are banned in the US and UK.</description><content:encoded><![CDATA[<p>My rule&#39;s blunt. If no real person stands behind a testimonial on your site, I&#39;d take it down, because a quote nobody can trace does more harm than the empty space it leaves.</p>
<p>I know that sounds harsh. Here&#39;s why I think it, and what I&#39;d put there instead.</p>
<h2>What visitors already assume about your quotes</h2>
<p>They start out doubting them. In Nielsen Norman Group&#39;s 2016 study of trust, people treated the quotes and case studies on a company&#39;s own site with caution: they wondered whether they were real, and they assumed a company would only ever show the nice ones. The researchers put it bluntly, saying &quot;people trust testimonials from external sites more than those listed on the website itself.&quot;</p>
<p>So a testimonial doesn&#39;t start at neutral. It starts a little behind, and every detail you give it either earns that back or proves the visitor right.</p>
<h2>Why vague praise reads as fake</h2>
<p>Because that&#39;s what fakes sound like, and people can&#39;t tell the two apart by reading them. In 2011, Myle Ott and colleagues at Cornell put 400 real hotel reviews beside 400 invented ones, written for a dollar each on Amazon Mechanical Turk by people told to pose as happy guests.</p>
<p>Three human judges tried to sort 160 of them. They scored between 53% and 62%, which the authors called &quot;roughly at-chance&quot;.</p>
<p>Detail is what gave the game away. The real reviews talked about the place itself (the size of the room, the bathroom, where it was), while the made-up ones drifted off to things outside the hotel, a husband or a business trip, and preened themselves on superlatives. That was hotels rather than software, and I wouldn&#39;t push the comparison too far. Still, you&#39;ll recognise the pattern on plenty of landing pages.</p>
<p>I see the same shape a lot in our audits. A row of quotes says &quot;Love it&quot;, &quot;Best tool ever&quot; or &quot;Five stars, would recommend&quot;, signed with a first name or no name at all. None of them has to be fake. A visitor just has no way of knowing, and the words don&#39;t give them anything to go on either way.</p>
<h2>The four things that make a testimonial believable</h2>
<p>A visitor should be able to check who said it, and learn something they couldn&#39;t have guessed. That comes down to four things.</p>
<ol>
<li><strong>A full name.</strong> A first name alone beats nothing, and some customers won&#39;t give more than that. A full name is something a visitor can actually look up.</li>
<li><strong>A role or company.</strong> It tells the visitor whether this person&#39;s like them. &quot;Head of growth at a 12-person agency&quot; places the quote, and &quot;Happy customer&quot; doesn&#39;t.</li>
<li><strong>A real photo.</strong> A face makes it feel like a person said it.</li>
<li><strong>A link to where they said it.</strong> Their own post, a public profile or a review on an outside site. This one deals with the doubt from the section above head on, since the visitor no longer has to take your word for anything.</li>
</ol>
<p>After that, the words need one concrete thing in them. What they used it for, what changed, or a number they&#39;d stand behind.</p>
<p>Here&#39;s a made-up plant care app at fernly.example, before and after:</p>
<table>
<thead>
<tr>
<th>Before</th>
<th>After</th>
</tr>
</thead>
<tbody><tr>
<td>&quot;Love it, best app ever.&quot; Sarah</td>
<td>&quot;I used to lose a fern every winter. Two winters in, all 11 of my plants are still alive.&quot; Sarah Okafor, nurse, Leeds, with her photo and a link to her own post</td>
</tr>
<tr>
<td>&quot;Fernly is amazing for our team.&quot; Mark</td>
<td>&quot;We look after the plants in three offices. Fernly cut our watering rounds from daily to twice a week.&quot; Mark Ellis, office manager at Brightdesk, with his photo</td>
</tr>
</tbody></table>
<p>The after versions gush less. They&#39;re also far easier to believe, and I&#39;d make that trade every single time, even though it means going back to customers and asking them for a sentence they have to think about.</p>
<h2>The photo is the part to be careful with</h2>
<p>In a 2012 set of experiments, Eryn Newman and colleagues showed people claims about unfamiliar names, some with a photo of the person and some without. The photos proved nothing either way. People still judged the claims true more often when there was one.</p>
<p>That cuts both ways. A real customer&#39;s photo helps their quote land, but a stock photo on a made-up quote borrows trust it hasn&#39;t earned, and I doubt a sceptical visitor who&#39;s seen that face somewhere else trusts anything near it again. If you haven&#39;t got their photo, use the name, role and link, and leave the picture out. An empty circle with their initials is fine.</p>
<h2>Where the law draws the line</h2>
<p>Making up a testimonial is now against the law in both the US and the UK. The FTC&#39;s rule on reviews and testimonials, in force since 21 October 2024, bars a business from writing or selling one that misrepresents that the person exists, that they used the product, or what their experience was. It also covers quotes a business spreads when it &quot;knew or should have known&quot; they were false, and ones it gets its own staff or their relatives to write that do the same.</p>
<p>The UK goes a step further. Under the Digital Markets, Competition and Consumers Act, whose rules on this came into force on 6 April 2025, the Competition and Markets Authority&#39;s guidance defines a fake review as one that &quot;purports to be, but is not, based on a person&#39;s genuine experience&quot;, and it warns against picking only the good ones &quot;when these do not reflect the experience being reported by reviewers overall&quot;. That applies to a business showing reviews on its own website. A quote that was paid for, given in return for a free product, or written by someone on your staff or a supplier needs to say so.</p>
<p>You can still pick your best quotes. They have to be real, and the set as a whole has to be fair to what customers actually say.</p>
<h2>Where to put them</h2>
<p>Next to the claim each one backs up, I think, though I should say I haven&#39;t found a study that tests where quotes ought to sit, so this is reasoning rather than evidence. A quote about saving time goes by the line that promises it. One about support goes near the pricing, where people worry about being left on their own once they&#39;ve paid. A grid of unrelated praise at the foot of the page leaves the visitor to do all that matching themselves, so get your ducks in a row for them.</p>
<p>On <a href="https://magpieaudit.com/articles/landing-page-first-screen">the first screen</a>, keep proof to one short line close to <a href="https://magpieaudit.com/articles/call-to-action-button-text">what the main button says</a>, either a single named quote or <a href="https://magpieaudit.com/articles/ratings-and-review-badges">a rating with its source</a>. The longer quotes can wait further down, where someone who&#39;s weighing it up will actually read them. A quote from a customer the visitor has heard of is a feather in your cap, by the way, and I&#39;d expect it to do more than a strip of <a href="https://magpieaudit.com/articles/customer-logos-on-landing-pages">customer logos</a> on its own.</p>
<p>If you&#39;d like an outside view of your own quotes, Magpie&#39;s full report checks each one your site shows for the four things above, and reads whether its words are specific and come from a customer rather than from the team behind the product.</p>
<h2>Questions people ask</h2>
<h3>How many testimonials should I have?</h3>
<p>Enough to back each main claim, and no more than real customers have given you. One specific, named quote does more than a dozen vague ones, so add them as they come in.</p>
<h3>Can I edit what a customer said?</h3>
<p>Trimming it for length and fixing a typo is fine, as long as it still says what they meant. What I wouldn&#39;t do is add a result they never mentioned, or make it sound keener than they were. And send them the version you&#39;re using so they can say yes to it, which is polite anyway and, if it ever comes up, means you can show they agreed to the exact words on your site.</p>
<h3>Are video testimonials better?</h3>
<p>They&#39;re harder to fake, I think, and there&#39;s a real face in them, so they carry trust well. They also ask the visitor for a minute or two. I&#39;d pair each one with a short written line from the same person and a link to the video, so a skimmer gets the point and a careful buyer can still watch the proof.</p>
]]></content:encoded></item>
<item><title>What your call to action button should say</title><link>https://magpieaudit.com/articles/call-to-action-button-text</link><guid isPermaLink="true">https://magpieaudit.com/articles/call-to-action-button-text</guid><pubDate>Tue, 17 Nov 2026 15:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Calls to action</category><category>First screen</category><description>Your call to action button should say what the visitor gets or what happens next, in a few plain words like &quot;Get my free report&quot; or &quot;See pricing&quot;. Labels such as &quot;Get started&quot;, &quot;Learn more&quot; and &quot;Submit&quot; fit any goal, so they mislead people who wanted something else. Keep one main button, and put the last doubt in a short line under it.</description><content:encoded><![CDATA[<p>Six of eight people on one meal delivery site pressed &quot;Get Started&quot; because they thought it would tell them about the company. It didn&#39;t. It dropped them into a sign-up flow, and that&#39;s from a study Nielsen Norman Group wrote up in 2017, which is still the clearest picture I know of what a vague button does to the people who press it.</p>
<p>So write the words on your button as if the visitor will hold you to them. Say what they&#39;ll get, or the step they&#39;re about to take, and use as few words as that needs.</p>
<h2>Why &quot;Get started&quot; and &quot;Learn more&quot; cost you</h2>
<p>They fit every goal. That&#39;s the trouble with them.</p>
<p>In the same study, some people on a second site saw &quot;Get Started&quot; sitting above the fold and decided the details they were after weren&#39;t on the page at all, so they never scrolled to find them. The researchers called the label &quot;ambiguous&quot; because it &quot;can apply to almost any goal that a user might have&quot;, and I&#39;d go a bit further. Even a sharp-eyed visitor can&#39;t tell from those two words whether a price, a menu or a form is waiting on the other side.</p>
<p>You still get the click in your analytics, of course. You just don&#39;t get much from someone who wanted a price and found a form.</p>
<p>&quot;Learn more&quot; has a second problem, and it&#39;s one most people never see. Someone using a screen reader often hears a page&#39;s links read out as a list, stripped of the sentences around them, so five links all saying &quot;Learn more&quot; turn into five identical sounds in a row (try telling those apart). WCAG 2.4.4, at level A, which is the lowest bar the accessibility guidelines set, asks that a link&#39;s purpose be clear from its own words, or from its words with the text right around them.</p>
<h2>Name what the visitor gets</h2>
<p>Use a verb and the thing itself, in the words a buyer would use. Take a made-up plant care app at fernly.example:</p>
<table>
<thead>
<tr>
<th>Vague</th>
<th>Says what happens</th>
</tr>
</thead>
<tbody><tr>
<td>Get started</td>
<td>Start watering reminders</td>
</tr>
<tr>
<td>Learn more</td>
<td>See how reminders work</td>
</tr>
<tr>
<td>Submit</td>
<td>Send me my plant plan</td>
</tr>
<tr>
<td>Sign up</td>
<td>Create my free account</td>
</tr>
<tr>
<td>Request a demo</td>
<td>Book a 20-minute walkthrough</td>
</tr>
</tbody></table>
<p>The right-hand column is longer, and I don&#39;t think that matters much. Each one tells you what you&#39;re agreeing to before you agree to it. It also sorts the clicks for you, because whoever presses &quot;Book a 20-minute walkthrough&quot; already knows a call is coming and isn&#39;t going to be annoyed when it does.</p>
<h3>Match the words to where the button leads</h3>
<p>If &quot;See pricing&quot; opens a contact form, the visitor notices. Fair enough if they leave. When the next step is a quiz, say quiz, and when it asks for an email, &quot;Email me the guide&quot; is honest about that. NN/g&#39;s fix for a fashion site in the same study went the same way, swapping &quot;Get Started&quot; for &quot;Take Our Style Quiz&quot;.</p>
<p>This means opening your own next page and reading it as a stranger would, which is easy to skip when you built both pages and know exactly where they go. Magpie&#39;s full report does it for you. It follows the main button one step (opening a link or pressing a plain button, never a form&#39;s) and reads the button&#39;s words against the page it lands on.</p>
<h2>One main button, and the same words each time</h2>
<p>&quot;Start free&quot; and &quot;Book a demo&quot; side by side, the same size and colour, make the visitor stop and pick. Some pick neither. I&#39;d make one of them the clear main button and turn the other into a plain link beside it.</p>
<p>Buttons that go to the same place should be birds of a feather, too. A header button saying &quot;Get started&quot; and a hero button saying &quot;Start free trial&quot; might lead to one page, but the visitor has no way of knowing that, so choose one label and use it in both. Repeating the main button in the header is pretty much always fine, by the way, since it keeps the action in reach as people scroll down.</p>
<p>The headline and the line under it matter as much as the button itself. I go through both in <a href="https://magpieaudit.com/articles/landing-page-first-screen">what a landing page&#39;s first screen should say</a>, with a made-up page to show each change.</p>
<h2>Put the last doubt in the line under it</h2>
<p>The few words under a button can do as much work as the button. Right at the moment of clicking, the doubt is usually about cost, a card, time, or what happens afterwards. Answer whichever one <a href="https://magpieaudit.com/articles/questions-your-landing-page-should-answer">your buyers ask most often</a>:</p>
<ul>
<li>No card needed</li>
<li>Takes about a minute</li>
<li>Cancel any time from your account</li>
<li>Free for up to five plants</li>
</ul>
<p>Only crow about &quot;No card needed&quot; if the sign-up really asks for none (whether <a href="https://magpieaudit.com/articles/free-trial-card-up-front">a free trial should ask for a card</a> at all is a separate argument). Someone who was ready to go and then hits a card form has just learned that your site says things that aren&#39;t true, and I doubt they&#39;ll read the rest of it kindly.</p>
<p>When the doubt is &quot;does this work for people like me&quot;, a <a href="https://magpieaudit.com/articles/landing-page-testimonials">named customer&#39;s quote</a> near the button does better than anything you write yourself.</p>
<h2>What colour should the button be?</h2>
<p>Whichever one stands out most from what&#39;s around it. I&#39;d love a firmer answer, but there isn&#39;t one. Andrew Elliot&#39;s 2015 review of colour research found the field still young, a lot of its studies too small, and every effect it did find tied to its context. What we do know is that a button catches the eye by standing apart from its surroundings, and no single colour has been shown to do that on its own, so pick yours for contrast and spend your testing time on the words. (Where eyes actually land is a separate thing, covered in <a href="https://magpieaudit.com/articles/where-visitors-look-first">where visitors look first</a>.)</p>
<p>Size matters more for thumbs than colour does. WCAG 2.5.8 puts the smallest tap target at 24 by 24 CSS pixels, and a main button on a phone wants to be comfortably bigger than that. Check it&#39;s still on the mobile first screen, too, since a tall picture above it is the usual reason it slips out of view.</p>
<h2>Questions people ask</h2>
<h3>Should it say &quot;my&quot; or &quot;your&quot;?</h3>
<p>Honestly, whichever reads more naturally on your site. The tests people usually quote for &quot;Start my trial&quot; over &quot;Start your trial&quot; are company case studies that don&#39;t publish a sample size, so I wouldn&#39;t choose on the strength of them. Saying what happens next matters far more than the pronoun.</p>
<h3>How long should the label be?</h3>
<p>I&#39;d aim for two to five words. If it won&#39;t fit in five, the extra detail probably wants to live in the line underneath instead.</p>
<h3>Is &quot;Get started&quot; ever fine?</h3>
<p>Sometimes, yes. If the words right around it already say exactly what starts, and the page is only offering one thing, I don&#39;t think it does much harm, though I&#39;ve not found a study either way. I&#39;d still swap it. Naming the thing (&quot;Start my free check&quot;) costs you nothing, it reads better to anyone on a screen reader, and it means the label still makes sense when someone copies your button into a different part of the site six months later and forgets what the paragraph above it used to say.</p>
]]></content:encoded></item>
<item><title>What a cookie banner on a landing page has to do</title><link>https://magpieaudit.com/articles/cookie-banners-on-landing-pages</link><guid isPermaLink="true">https://magpieaudit.com/articles/cookie-banners-on-landing-pages</guid><pubDate>Tue, 17 Nov 2026 15:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>First screen</category><description>Under UK and EU rules, a cookie banner holds back every cookie, pixel and tag that isn&apos;t essential until the visitor says yes, offers Reject as plainly as Accept, and lets them change their mind later. Since February 2026, UK sites may count visits unasked if the figures stay theirs and visitors can opt out. On a phone, a slim bottom bar keeps your headline in view.</description><content:encoded><![CDATA[<p>A cookie banner has two jobs in the UK and the EU. It holds back everything that isn&#39;t essential until the visitor agrees, and it makes saying no as easy as saying yes. The size, the place and the wording are yours to choose, and that&#39;s where most of the harm to a first screen gets done.</p>
<p>The first job is the one that fails quietly. The page looks fine, the banner sits there politely, and the advertising pixels have already fired behind it.</p>
<h2>What UK and EU rules ask of a cookie banner</h2>
<p>Nothing that needs consent may run until the visitor gives it. The ICO&#39;s final guidance on storage and access technologies, published on 29 April 2026, says that where no exception applies &quot;you must obtain prior consent&quot;, and that silence or a visitor who ignores the banner doesn&#39;t count as a yes. It reaches past cookies, too. The ICO lists tracking pixels, web storage, fingerprinting and &quot;scripts and tags&quot;, so an advertising pixel that sends data without setting a cookie is still in scope.</p>
<p>Then comes how the choice is offered. The ICO&#39;s checklist wants a banner that &quot;makes it as easy to refuse consent as it is to accept&quot;, and its good example puts &quot;accept all&quot; and &quot;reject all&quot; side by side and equally prominent, with any finer switches off until the visitor turns them on. Its bad example has Accept and no way to reject everything at once.</p>
<p>The EU&#39;s regulators say much the same. In the EDPB&#39;s cookie banner taskforce report, adopted on 17 January 2023, &quot;a vast majority&quot; of the data protection authorities treated a banner with no way to refuse, on any layer that has a consent button, as an infringement (a few disagreed, as the ePrivacy Directive doesn&#39;t spell out a reject option). The same report says pre-ticked boxes don&#39;t give valid consent, and nor does a Reject tucked into a paragraph as a link with nothing to draw the eye to it. France&#39;s CNIL sent formal notices over exactly that faint link in December 2024, saying &quot;rejecting cookies should be just as easy as accepting them&quot;.</p>
<p>Last, a way back. The ICO says visitors must be able to withdraw consent at any time &quot;with the same ease that they gave it&quot;, and the EDPB suggests a small icon or a link in a standard place, which for most sites means the footer.</p>
<p>I&#39;m not a lawyer and this isn&#39;t legal advice. Here are the four side by side, from the regulators&#39; own pages.</p>
<table>
<thead>
<tr>
<th>The banner has to</th>
<th>ICO, UK (2026)</th>
<th>EDPB, EU (2023)</th>
</tr>
</thead>
<tbody><tr>
<td>Hold back what needs consent</td>
<td>Prior consent; silence isn&#39;t consent</td>
<td>No cookie needing consent set without it</td>
</tr>
<tr>
<td>Offer Reject as plainly as Accept</td>
<td>Equally prominent accept all and reject all</td>
<td>No refuse option on any layer is an infringement</td>
</tr>
<tr>
<td>Leave the choices off until a yes</td>
<td>Finer switches off by default</td>
<td>Pre-ticked boxes aren&#39;t valid consent</td>
</tr>
<tr>
<td>Let visitors change their mind</td>
<td>Withdraw as easily as they gave it</td>
<td>An icon or a link in a standard place</td>
</tr>
</tbody></table>
<p>The ICO has been checking. In December 2025 it said 979 of the UK&#39;s 1,000 busiest websites passed its tests, 564 of them only after failing first, and in April 2026 it put the figure at 99%. Those are big sites with lawyers. When we visited sixteen software and shop sites in a headless browser from a UK address in November 2026, three set advertising cookies while their banner was still waiting for an answer, and one carried on after Reject was pressed. Sixteen is a thin sample, so treat it as a hint and check your own. If this part is wrong, no amount of banner design will put it right.</p>
<h2>Whether your analytics need a banner</h2>
<p>In the UK, possibly not any more. The Data (Use and Access) Act 2025 added an exception to PECR for statistics, fully in force since 5 February 2026. Analytics can run without asking when they only count how your site is used, the figures aren&#39;t shared except with whoever helps you work them out, visitors are told what it&#39;s for, and they get a simple, free way to object.</p>
<p>The ICO reads that narrowly. An analytics company can do the counting for you, but only as your processor, and once the data is linked to adverts it needs consent again (&quot;Online advertising purposes are not exempt from PECR&#39;s consent requirements and never have been&quot;). Total visits, average scroll depth, A/B tests, load speed and bounce rates are on its list of what&#39;s likely covered, while recordings of individual visitors and what they did are not. So a Meta pixel, a LinkedIn tag, a session recorder or analytics feeding your ad campaigns still need a yes first.</p>
<p>If a landing page only needs to know how many people came and which section they reached, and nothing else on it needs consent, I&#39;d set the analytics up to do just that and drop the banner for UK visitors. You&#39;d still need a line in your privacy policy (one that <a href="https://magpieaudit.com/articles/site-promises-and-your-terms">matches what your site promises</a>) saying what&#39;s counted and why, and an opt-out a visitor can actually find, such as a footer link to a page with a switch. The ICO doesn&#39;t fix the form, though it says browser settings alone won&#39;t do.</p>
<p>Elsewhere in Europe it depends on the regulator. France&#39;s CNIL has let audience measurement run unasked since 2020, on its own conditions: one publisher only, the last part of the IP address dropped, trackers kept 13 months at most, and visitors told and able to object. It also says &quot;Most large audience measurement offerings do not fall within the scope of the exemption, regardless of their configuration&quot;, which is worth reading twice before you decide your setup qualifies.</p>
<h2>What a cookie banner costs your first screen</h2>
<p>On a phone it can take a big share of it. Google&#39;s web.dev guide says a cookie notice &quot;typically takes up a larger portion of the screen&quot; on mobile, sometimes enough that its own text becomes the largest thing the page draws, which is the element your load time is measured by.</p>
<p>Here&#39;s a made-up seed subscription shop at sproutbox.example, on a 393 by 852 pixel phone screen (an iPhone 14 Pro&#39;s).</p>
<table>
<thead>
<tr>
<th></th>
<th>Before</th>
<th>After</th>
</tr>
</thead>
<tbody><tr>
<td>Where it sits</td>
<td>A card in the middle of the screen</td>
<td>A bar along the bottom</td>
</tr>
<tr>
<td>Height</td>
<td>380px, 45% of the screen</td>
<td>150px, 18% of the screen</td>
</tr>
<tr>
<td>Words</td>
<td>Three paragraphs about partners and purposes</td>
<td>One sentence and a link to the full policy</td>
</tr>
<tr>
<td>Buttons</td>
<td>Accept all, Manage options</td>
<td>Accept and Reject, the same size and style</td>
</tr>
<tr>
<td>What it covers</td>
<td>The line under the headline and the button</td>
<td>The bottom 150px of the screen only</td>
</tr>
</tbody></table>
<p>The before version also breaks the rules above, since Reject sits a layer down. The after version leaves the headline, the line under it and the Start my box button where <a href="https://magpieaudit.com/articles/landing-page-first-screen">the first screen</a> needs them.</p>
<p>What it costs in sign-ups, I can&#39;t tell you. I haven&#39;t found a study that measures it on landing pages. The web.dev guide warns that &quot;smaller sites may find that users bounce rather than accept the cookie notice of an unfamiliar site with obscured content&quot;, which is Google&#39;s advice, not a measurement.</p>
<p>What has been measured is how a banner&#39;s design steers the answer. Nouwens and colleagues (CHI 2020) scraped the five most popular consent tools on the UK&#39;s top 10,000 sites and found only 11.8% of the 680 banners met the minimal legal requirements they set. In their experiment with 40 people, taking Reject off the first page raised consent by 22 to 23 percentage points. That&#39;s the gain a hidden Reject buys, and it&#39;s the design the ICO and the EDPB call out. Utz and colleagues (CCS 2019) tested notices on a German site with more than 80,000 visitors and found people were more likely to deal with one shown in the lower part of the screen, the left especially. That&#39;s handy for a landing page, because the bottom of a phone screen is the part your headline doesn&#39;t need, and a banner that gets answered goes away.</p>
<p>A banner can also make a page feel slow. Web.dev calls cookie notices &quot;a very common source of layout shifts&quot; when one is slotted in at the top and pushes everything down, and pressing Accept can load a pile of other sites&#39; scripts at once and stall the next tap. A visitor reaching for your button as the page lurches is a sitting duck. The fixes are the same ones in <a href="https://magpieaudit.com/articles/landing-page-speed-on-mobile">how fast a landing page should load on a phone</a>.</p>
<h2>A cookie banner that does its job and leaves the headline alone</h2>
<p>This is what I&#39;d put on a landing page, from the rules and the guidance above.</p>
<ul>
<li>A bar along the bottom, drawn over the page so nothing moves when it appears. Eyes go to the top left first (<a href="https://magpieaudit.com/articles/where-visitors-look-first">where visitors look first</a>), so that spot stays your headline&#39;s.</li>
<li>One plain sentence on what you&#39;d like to use and why, with a link to the cookie policy for the detail.</li>
<li>Accept and Reject with the same size, colour and weight, and a quieter Settings link beside them if you offer finer choices.</li>
<li>Advertising, recording and anything else that needs a yes loaded only after Accept, from the banner&#39;s own answer, never from the page&#39;s load.</li>
<li>The banner&#39;s script loaded asynchronously and straight from your page rather than through a tag manager, as web.dev suggests, and a Cookie settings link in your footer for changing your mind.</li>
</ul>
<p>I wouldn&#39;t agonise over the wording, as long as both buttons say what they do. &quot;Accept&quot; and &quot;Reject&quot; are fine. A bright Accept beside a faint &quot;continue without accepting&quot; link is the pattern the EDPB and the CNIL called out.</p>
<h2>How to check what your site runs before anyone answers</h2>
<p>Open your site in a private window from a UK or EU connection, and before you touch the banner, open the browser&#39;s developer tools. Look at the cookies the page has set and at the network requests it&#39;s made, searching for names such as facebook, doubleclick, linkedin, clarity or hotjar. Anything advertising or recording that&#39;s already there ran before consent. Then press Reject, reload and look again.</p>
<p>Where you check from matters, because some banner tools show visitors outside the UK and EU no banner at all and load everything. A check from New York can make a careful site look careless.</p>
<p>Magpie&#39;s full report does this in code on its phone visit. A tracker counts as running only when it sets its cookie or sends data (a script that loads and waits doesn&#39;t), and what&#39;s seen is split at the moment the banner is answered, so the report shows what ran before anyone said yes and what still ran after Reject. Consent is judged only when the visit comes from the UK, the EU or the EEA and the banner was seen, and the report also checks that your privacy policy names each tracker&#39;s company. For now our visits come from the US, so the report lists what runs and doesn&#39;t yet judge consent.</p>
<h2>Questions people ask</h2>
<h3>Will a full-screen cookie banner hurt my Google ranking?</h3>
<p>Not by itself. Web.dev says &quot;Google Search does not penalize the usage of interstitials when they are used to comply with legal regulations such as in the case of cookie banners&quot;. Visitors are another matter, and few of them are happy as larks about a screen they have to clear before they can read anything.</p>
<h3>Do I need a banner if my site sets no cookies?</h3>
<p>It depends on what else the page does, since the rules cover pixels, local storage, fingerprinting and scripts as well as cookies. If the only things stored are strictly necessary (the ICO&#39;s examples include a shopping basket, a logged-in session and security for a service someone asked for), you need no consent and no banner. If your analytics fit the UK&#39;s statistics exception, you need the explanation and the opt-out but not the question. The moment an advertising pixel or a session recorder goes on, you&#39;re back to asking first, and &quot;we don&#39;t use cookies&quot; stops being the test.</p>
]]></content:encoded></item>
<item><title>How to get your product named in AI answers</title><link>https://magpieaudit.com/articles/named-in-ai-answers</link><guid isPermaLink="true">https://magpieaudit.com/articles/named-in-ai-answers</guid><pubDate>Tue, 17 Nov 2026 12:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>Search and AI</category><description>AI answers name the products their web search finds and can read. Let each assistant&apos;s search crawler in, send your key sentences in the HTML rather than drawing them with scripts, say plainly what the product is, who it&apos;s for and what it costs, and get written about on other sites. Rewording pages with GEO tricks did little or harm in a 2025 benchmark.</description><content:encoded><![CDATA[<p>You asked ChatGPT for the best tool in your category and it named four of your rivals. Not you. That stings, and I think it&#39;ll matter more each year. Pew Research tracked 900 US adults&#39; searches in March 2025 and found that when Google showed an AI summary, people clicked a normal result on 8% of visits, against 15% when there wasn&#39;t one.</p>
<p>The assistants are a lot less mysterious than the people selling &quot;GEO&quot; make them sound. They run a web search, read what comes back and write their answer from it. So there are two things worth finding out about your site, and neither needs a consultant. Does their search find it? And once it has, can it read what&#39;s there?</p>
<h2>Why an assistant leaves your product out</h2>
<p>Google is blunt about its side. A page in AI Overviews or AI Mode must be &quot;indexed and eligible to be shown in Google Search with a snippet&quot;, and both features may run &quot;multiple related searches across subtopics and data sources&quot; to build a single answer. The others send search crawlers of their own (OAI-SearchBot for ChatGPT, PerplexityBot for Perplexity, Claude-SearchBot for Claude), and those fetch your pages to answer the searches their assistant runs.</p>
<p>So you can be missed in three ways. The crawler was turned away at the door. Or it got in and found an empty shell where your words should&#39;ve been. Or it read you perfectly well and other pages simply ranked higher for that search, which is the slow one, and mostly ordinary search work, right down to what <a href="https://magpieaudit.com/articles/google-title-and-description">your Google title and description</a> say.</p>
<p>The first two you can often fix in an afternoon.</p>
<h2>Let each assistant&#39;s search crawler in</h2>
<p>I&#39;d guess most sites that block AI crawlers meant to block training and caught search by accident. Each company runs separate crawlers for the two jobs, and a robots.txt rule written in a hurry (say, a list of every AI user agent copied from a forum post) shuts you out of the answers as well as the training data.</p>
<table>
<thead>
<tr>
<th>Assistant</th>
<th>Search crawler</th>
<th>What its owner says about blocking it, as of November 2026</th>
</tr>
</thead>
<tbody><tr>
<td>ChatGPT</td>
<td>OAI-SearchBot</td>
<td>Sites opted out &quot;will not be shown in ChatGPT search answers, though can still appear as navigational links&quot;</td>
</tr>
<tr>
<td>Perplexity</td>
<td>PerplexityBot</td>
<td>It is &quot;designed to surface and link websites in search results&quot; and &quot;not used to crawl content for AI foundation models&quot;</td>
</tr>
<tr>
<td>Claude</td>
<td>Claude-SearchBot</td>
<td>Blocking it may reduce your &quot;visibility and accuracy in user search results&quot;</td>
</tr>
<tr>
<td>Google AI Overviews</td>
<td>Googlebot</td>
<td>The same crawler as Google Search, so blocking it takes you out of both</td>
</tr>
</tbody></table>
<p>Training is a separate choice. OpenAI says disallowing GPTBot means your content &quot;should not be used in training&quot;, and that it&#39;s an independent setting from OAI-SearchBot. You can keep your words out of training and still turn up in ChatGPT&#39;s search, and I think that&#39;s the right default for most products.</p>
<p>Check your CDN as well. Since 1 July 2025 Cloudflare has asked every new domain at sign-up whether to let AI crawlers in, so a site can be shut to them at the edge without a single line in robots.txt. If you clicked no in a hurry while sorting out DNS, you&#39;ve locked your site in a birdcage. You can change it later in the dashboard.</p>
<h2>Send your words in the HTML, not only by script</h2>
<p>In December 2024 Vercel and MERJ published a study of AI crawler traffic and reported that &quot;none of the major AI crawlers currently render JavaScript&quot;. GPTBot, ClaudeBot and PerplexityBot fetch the script files and never run them. Googlebot does render pages, which is how a site can look fine in Google and be blank to ChatGPT (and why <a href="https://magpieaudit.com/articles/shared-link-previews">a shared link&#39;s card</a> can come up empty too).</p>
<p>Take shiftjar.example, a made-up rota app for cafés. Open it in a browser and the first screen says &quot;Rotas for cafés, done in ten minutes a week&quot;. Ask its server for the page, though, and you get one empty <code>div</code> and a script. A crawler that runs no scripts reads nothing there at all, so no answer to &quot;rota app for cafés&quot; can name it, however good the product is. Rendering that first screen on the server (any framework&#39;s static or server rendering will do) puts the sentence where every crawler can read it.</p>
<p>Then give it something to quote. Someone asking for the &quot;best rota app for a small café&quot; wants a line saying what the product is, who it&#39;s for and what it costs, and a slogan gives the assistant nothing to lift. That&#39;s one more reason for <a href="https://magpieaudit.com/articles/show-prices-on-landing-page">showing prices on a landing page</a>, and the plain words are what <a href="https://magpieaudit.com/articles/landing-page-first-screen">a landing page&#39;s first screen</a> should open with anyway, for people and assistants alike.</p>
<p>Magpie&#39;s full report checks this part. It reads your page as the server sends it and says how many of its words arrive without scripts, reads your robots.txt for each assistant&#39;s search crawler, and picks the sentence an assistant could quote for five questions (what it is, who it&#39;s for, how it works, how it&#39;s different and what it costs). It also runs buyers&#39; searches and says whether today&#39;s answer names you.</p>
<h2>What GEO tricks do, according to the research</h2>
<p>C-SEO Bench is the paper I&#39;d send to anyone selling GEO. Accepted at NeurIPS&#39;s datasets and benchmarks track in 2025, it took eight rewriting methods from the original GEO paper (adding statistics, quotes, citations, an authoritative tone, technical terms and so on) and ran them across six kinds of search (from retail and books to news and debate) on four models, gpt-4o-mini, o3, o4-mini and Claude 3.5 Haiku. Out of 54 cases, three ranked better by a margin that held up statistically.</p>
<p>Three. And the Statistics method, which sounds like the safest of the lot, lowered rankings in 19 of 24 settings.</p>
<p>What did move a product up was sitting high in the search results the model read. On one retail test with gpt-4o-mini, putting a page first in the model&#39;s results raised its place in the answer by 2.77 positions on average, against 0.36 for the best rewriting method. The authors also found that the early adopters&#39; gains &quot;decrease steadily as more players adopt the same method&quot;, and called it &quot;both a congested and a zero-sum game&quot;.</p>
<p>To be fair to the 2024 GEO paper that started all this, it&#39;s a real paper, accepted at KDD, and it reported visibility gains of up to 40%. It tested engines and a benchmark from 2023, though, and the newer benchmark didn&#39;t find those gains on the models it tried. Models change fast enough that I hold both results a bit loosely.</p>
<p>Google&#39;s own guidance ends up in much the same place. You &quot;don&#39;t need to create new machine readable files, AI text files, or markup&quot;, there&#39;s &quot;no special schema.org structured data&quot; for AI features (so FAQ markup isn&#39;t a way in), and there&#39;s &quot;no requirement to break your content into tiny pieces&quot;. I wouldn&#39;t spend a day on an llms.txt file, or on chopping pages into question-and-answer chunks.</p>
<h2>Get written about on other sites</h2>
<p>A 2025 preprint by Chen, Wang, Chen and Koudas compared the sources AI search cites with Google&#39;s results. It isn&#39;t peer reviewed yet, so treat it as a lead rather than a verdict. It found &quot;a systematic and overwhelming bias towards Earned media&quot;, meaning reviews, articles and comparisons on other sites, over a brand&#39;s own pages and social posts. My reading, and it&#39;s only a reading, is that a &quot;best tools for&quot; question gets answered largely from best-of lists.</p>
<p>Which makes the slow work getting onto those lists. Comparison articles line a category&#39;s products up like birds on a wire, and one that names your four rivals and leaves you out is a likely place the assistant learned its list in the first place. Getting into the review sites, directories and comparison pieces in your category (with your own site saying the same plain sentences they&#39;ll quote) is the part no line in robots.txt can do for you.</p>
<h2>How to see whether answers name you</h2>
<p>Ask the same question several ways, on more than one assistant, on more than one day. The searches behind an answer shift with the wording and the date, so a single &quot;it named me&quot; (or didn&#39;t) tells you very little. And ask what a buyer would type, something like &quot;best rota app for a small café&quot;, rather than your brand name, which any assistant finds far more easily than it finds your category.</p>
<p>Microsoft&#39;s side has real data now. On 10 February 2026 Bing Webmaster Tools added an AI Performance report, in public preview, that counts how often your pages are cited as sources in Copilot and Bing&#39;s AI summaries, and shows a sample of the searches the AI ran when it cited them. It counts citations, not where in an answer you appear. Still, it&#39;s free, and it&#39;s the nearest thing I know of to a search console for AI answers.</p>
<h2>Questions people ask</h2>
<h3>Is an llms.txt file worth adding?</h3>
<p>Not for Google, which says Google Search doesn&#39;t use them. Other services might read one and it won&#39;t do any harm, so if it takes you five minutes, fair enough. I just wouldn&#39;t expect it to change what anyone&#39;s assistant says about you.</p>
<h3>Should I block GPTBot?</h3>
<p>That one&#39;s really about training rather than being found, and it&#39;s easy to mix the two up because the names look so alike. Blocking GPTBot keeps your words out of OpenAI&#39;s training. Whether you turn up in ChatGPT&#39;s search is down to OAI-SearchBot, which is a separate setting, so you can shut one and leave the other open. I said earlier that I&#39;d block training and leave search alone, and on most product sites that&#39;s still what I&#39;d do, though I can see someone deciding their marketing copy is something they&#39;d quite like repeated everywhere and leaving both open. Either is a fair call. The one I&#39;d argue against is blocking the search crawlers.</p>
]]></content:encoded></item>
<item><title>What a landing page&apos;s first screen should say</title><link>https://magpieaudit.com/articles/landing-page-first-screen</link><guid isPermaLink="true">https://magpieaudit.com/articles/landing-page-first-screen</guid><pubDate>Tue, 17 Nov 2026 09:00:00 GMT</pubDate><dc:creator>Alex</dc:creator><category>First screen</category><category>Calls to action</category><description>A landing page&apos;s first screen should say what the product does for the visitor, who it&apos;s for, and what to do next. The headline names the outcome in the buyer&apos;s own words, the line under it adds what the headline leaves out, and one button says what happens when it&apos;s pressed. Everything else can wait further down the page.</description><content:encoded><![CDATA[<p>Before anyone scrolls, your first screen has to tell them what the product does for them, who it&#39;s for, and what to press. If it manages those three, the rest of the page gets a chance.</p>
<p>I&#39;d take a plain first screen that does all three over a beautiful one that does one of them. Someone who has to work out what you sell mostly won&#39;t bother, and the numbers below show how little time you get to explain it before they don&#39;t.</p>
<h2>How much of the page gets seen</h2>
<p>In a 2018 eye tracking study of 120 people, Nielsen Norman Group found that 57% of viewing time went on the first screen and 74% on the first two. Back in 2010 the first screen alone took 80%.</p>
<p>So people do scroll more than they used to. They just need the top to give them a reason.</p>
<p>And they decide quickly. Jakob Nielsen&#39;s 2011 analysis of how long visits last found that the first 10 seconds are the critical ones for whether people stay, and his advice was to get your value across inside them. Plenty fly the nest within 10 to 20 seconds. The ones who stay don&#39;t read much either. They get through at most 28% of the words on an average visit, and more likely about 20%, which Nielsen worked out in 2008 from visit lengths at 250 words a minute (so it&#39;s an estimate from time spent, not from watching eyes, and I&#39;d treat the exact figure loosely).</p>
<h2>The headline names what the visitor gets</h2>
<p>A stranger should be able to repeat your headline back to you as a thing they&#39;d get. I think most weak ones fail because they name the category, or a slogan, or the technology underneath.</p>
<p>Here are three for a made-up plant care app at fernly.example:</p>
<table>
<thead>
<tr>
<th>Headline</th>
<th>What a stranger hears</th>
</tr>
</thead>
<tbody><tr>
<td>The future of plant care</td>
<td>Something about plants, maybe</td>
</tr>
<tr>
<td>Intelligent hydration scheduling</td>
<td>Software, for something wet</td>
</tr>
<tr>
<td>Know when each plant needs water</td>
<td>A watering reminder, per plant</td>
</tr>
</tbody></table>
<p>The last one&#39;s the dullest to write. It&#39;s also the one I&#39;d put up, because it uses the words a buyer actually types into a search box. Save the clever line for lower down, once the visitor knows what they&#39;re looking at.</p>
<p>One idea, too. A headline that promises reminders and a plant identifier and a community forum leaves the reader holding none of them.</p>
<h2>The line under it answers the next question</h2>
<p>Most subheadings say the headline again in longer words. That&#39;s a waste of the line a visitor is most likely to read next, and I think it&#39;s the cheapest fix on most first screens. Use it for whatever the headline makes them wonder, which is usually how it works or what it costs, and now and then who it&#39;s for.</p>
<p>For Fernly it might be: &quot;Snap a photo of each plant and get a reminder on your phone the day it needs watering. Free for up to five plants.&quot; That&#39;s 24 words, and the visitor now knows the effort involved (one photo each), where the reminder turns up (their phone) and that trying it costs nothing at all.</p>
<p>Only room for one of those? Pick the doubt your buyers bring up most, and if you&#39;re not sure which it is, your support inbox and your sales calls will settle it fast (I go through <a href="https://magpieaudit.com/articles/questions-your-landing-page-should-answer">finding the questions buyers really ask</a> separately).</p>
<h2>One button, saying what happens next</h2>
<p>Put one main button on the first screen, and give it words that say what pressing it does. &quot;Submit&quot; or &quot;Learn more&quot; could sit on any site on the web, while &quot;Start watering reminders&quot; could only belong to Fernly. There&#39;s more on <a href="https://magpieaudit.com/articles/call-to-action-button-text">what the button should say</a> in its own article.</p>
<p>Two buttons at equal weight split the decision. Make one the main action and drop the other back to a plain link, so the order (headline, line, button) reads at a glance.</p>
<p>Then look at it on a phone. A tall hero picture, or <a href="https://magpieaudit.com/articles/cookie-banners-on-landing-pages">a cookie banner</a> sitting over the bottom half, can push the button below the fold on mobile while it looks fine on the laptop you built the page on, and you&#39;d never know unless you checked. That&#39;s why Magpie&#39;s free check measures the main button on both screens, a 14-inch MacBook Pro&#39;s and an iPhone 14 Pro&#39;s.</p>
<p>The few words under the button are worth more than another feature would be. &quot;No card needed&quot; or &quot;Cancel any time&quot; answers the last doubt right when it bites. Only write it if it&#39;s true.</p>
<h2>What can wait further down</h2>
<p>Most of what you&#39;re proud of, to be honest. These can nearly always go on the second screen or later:</p>
<ul>
<li>the full feature list</li>
<li>the story of why you built it</li>
<li>long testimonials</li>
<li>a pricing table with every plan</li>
</ul>
<p>Proof is the exception, a bit of it anyway. One line near the button, either <a href="https://magpieaudit.com/articles/ratings-and-review-badges">a rating with its source</a> or <a href="https://magpieaudit.com/articles/landing-page-testimonials">a named customer&#39;s quote</a>, answers &quot;can I trust this?&quot; while the visitor is still deciding. A whole flock of logos nobody recognises is just more for a visitor to read before they reach the button.</p>
<h2>Questions people ask</h2>
<h3>Does everything important have to be above the fold?</h3>
<p>No, and trying to squeeze it all in usually makes things worse. Nielsen Norman Group&#39;s advice is to keep the key goals and the main call to action at the top and let everything else follow. I&#39;d think of the first screen as the bit that earns the scroll, because a page crammed into one screen tends to give people no reason to keep going, and the second and third screens are where the people who are actually interested do their reading anyway.</p>
<h3>Should my product&#39;s name be the headline?</h3>
<p>Not on its own. Your logo&#39;s already saying it, a few centimetres to the left, so spend the headline on what the product does.</p>
<h3>How can I tell if mine works?</h3>
<p>Show it to someone who&#39;s never seen it for five seconds, then ask what it is, who it&#39;s for and what they&#39;d press. If they get all three, you&#39;re done. If they get two, the missing one is your next edit.</p>
]]></content:encoded></item>
</channel>
</rss>