Broken links on landing pages: which lose visitors. No wild goose chases

· Founder of Magpie AI

Published · 8 min read

In short

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 "#" 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.

Contents

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 "Book a class" button in the header still points at /classes, so everyone who presses it lands on a 404 that says "Not Found" and nothing else.

Nobody at the studio sees it, because they type the new address from memory. The bookings just go quiet.

Most broken links aren't that bad. A dead link to a 2019 blog post in the footer costs you very little. The ones I'd fix today are those a visitor meets while they're trying to do the thing your site is for, which on a landing page means the main button, the menu and anything on the first screen.

Six kinds of dead end, roughly from worst to mildest:

What the visitor pressesWhat they meet
A link to a page that's goneA 404 or 410 page, often with no way forward
A link to a domain that's lapsedThe browser saying the site can't be reached
A button or "#" link with no scriptNothing at all, so they press it again, then leave
A link to your own broken pageA server error, the same on a second try
A jump to a section that was removedNothing at all; the page stays where it is
A link to one of your own pagesYour home page, with no word of why

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't tell whether it's your site, their signal or their thumb.

A link written as href="#" or javascript:void(0) relies on a script to do its job, and when the script hasn't loaded, has an error or was taken out in the redesign, the link goes nowhere. MDN calls these "fake buttons", and says they misbehave when someone opens the link in a new tab or bookmarks it, or while JavaScript "is loading, errors, or is disabled". Its advice is a real <button> for actions, and a link only "for navigation to a real URL".

On a phone this tends to show up in the menu. A burger menu whose "Pricing" item was wired to an old script is dead on mobile and fine on the laptop where you tested it.

A jump to a section that isn't there#

A link like /#pricing scrolls to the part of the page with id="pricing". Remove that section, or rename its id, and the HTML Standard's rule is that the fragment "does not identify anything", so the browser simply doesn't scroll. No error, no 404. The page just sits there while the visitor wonders why "See pricing" did nothing.

Your server never hears about it, either. As MDN puts it, the fragment "is not sent to the server when the URI is requested", so no server log or 404 report will ever count these. You only find them by pressing them.

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'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.

Google takes the same view. Its site move guide warns against sending many old addresses to "one irrelevant single URL destination, such as the home page", because it "can confuse users and might be treated as a soft 404 error".

How to find them#

Press everything, on a phone and on a laptop, including the menu once it's opened on the phone. That's the honest answer, and it's dull enough that I doubt many people do it after every release.

A link checker helps with the ones that have an address. Jakob Nielsen's advice in 1998 was to "run a link validator on your site at regular intervals", and it still holds. What a crawler of that kind can'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.

For addresses people arrive at from outside (old adverts, emails, other sites' links), Google Search Console's page indexing report lists the ones Google found under "Not found (404)" and "Soft 404", the 404s from the past month only. Your server's logs show which missing addresses people actually request, and how often. Neither shows you a dead "#" button.

Magpie'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's not found, a site that doesn't exist, a server error from your own site twice in a row, a "#" or button that does nothing when pressed, a jump to a part the page doesn't have, and a link to your own page that lands on the home page. Anything it can'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.

What to do with the old addresses after a redesign#

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've landed.

Google's guide to redirects says to use "a permanent server-side redirect whenever possible", and names 301 and 308 as the codes that mean a page "has permanently moved to a new location". Its crawler documentation calls 308 "equivalent to 301". A temporary 302 or 307 keeps the old address in search results, which is fine for a page that's coming back and wrong for one that isn't.

For kilnhouse.example the map is short:

Old addressSend it toHow
/classes/workshops301
/classes/wheel-101/workshops/beginners-wheel301
/gift-vouchers/shop/vouchers301
/summer-2024-campNothing new; it's over410, with help
/#timetable/workshops#timetableFix the link

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'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 "ideally no more than 3 and fewer than 5" hops. And "keep the redirects for as long as possible, generally at least 1 year".

The last row in that table is a jump, so no redirect can help it (the server never sees the #timetable part). Search your own templates for links to ids that no longer exist and change them by hand.

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'd rather keep the address alive and point it at a follow-up or the nearest current page. I'd lean that way too for anything that was ever linked from an advert or a newsletter.

What your 404 page should say#

Say plainly that the page isn't here, then offer the likeliest ways on. NN/g's 2023 error message guidelines single out the 404 as the web's most common error message and one that breaks their rule on plain words, and they ask every error to "offer some potential remedies". Nielsen'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.

On a small marketing site I'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's "This page has moved or gone. Most people looking for it want" followed by Workshops, Gift vouchers and the timetable.

Make sure it answers with a real 404 status, too. A "page not found" 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.

Not on their own. Google's Susan Moskwa wrote in 2011 that if some URLs on your site 404, "this fact alone does not hurt you or count against you in Google's search results". 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're found is covered in your Google title and description.)

Questions people ask#

Should I use a 301 or a 308?#

Either. Google treats them the same. Some servers and hosting platforms only offer one, and that's fine.

After every release that touches the menu, the main button or page addresses, and I'd say about once a month otherwise. I'd add one more moment that'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'll only hear about it when someone emails to say the "Book now" button on your pricing page goes to a page that doesn't exist.

Less than your own, unless they'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.

Sources

  1. Google Search Central, Redirects and Google Search (opens in a new tab)
  2. Google Search Central, Site moves with URL changes (opens in a new tab)
  3. Google Search Central, Troubleshoot Google Search crawling errors (soft 404s) (opens in a new tab)
  4. Google, How HTTP status codes affect Google's crawlers (opens in a new tab)
  5. Google Search Console Help, Page indexing report (opens in a new tab)
  6. Google Search Central Blog, Do 404s hurt my site? (Susan Moskwa, 2011) (opens in a new tab)
  7. Nielsen Norman Group, Fighting Linkrot (Jakob Nielsen, 1998) (opens in a new tab)
  8. Nielsen Norman Group, Improving the Dreaded 404 Error Message (Jakob Nielsen, 1998) (opens in a new tab)
  9. Nielsen Norman Group, Error-Message Guidelines (Tim Neusesser and Evan Sunwall, 2023) (opens in a new tab)
  10. MDN, The Anchor element (opens in a new tab)
  11. MDN, URI fragment (opens in a new tab)
  12. WHATWG, HTML Standard: scrolling to a fragment (opens in a new tab)

About the author

· Founder of Magpie AI

Alex founded Magpie AI and writes about what turns a landing page's visitors into customers, drawing on what its audits find on real sites, from the first screen to the checkout

Related articles

Fix what loses you customers.Fix what loses you customers. Then bring more of them in

  • Free score, no sign‑upNo sign‑up
  • No credit card neededNo card needed