How fast a landing page should load on a phone. Before visitors take flight

Alex · Founder of Magpie AI
Published · 9 min read

In short
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'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' scripts later.
Contents
- The three numbers Google measures
- Why your phone score is worse than your laptop's
- How to read a red PageSpeed Insights result
- Does speed change how many visitors sign up?
- Does page speed affect Google ranking?
- The fixes that matter most on a landing page
- What Magpie's free check measures
- Questions people ask
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.
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.
The three numbers Google measures#
They're called Core Web Vitals, and each one times a different way a page can feel slow. These are web.dev's thresholds.
| Measure | What it times | Good | Poor |
|---|---|---|---|
| Largest Contentful Paint (LCP) | When the largest picture, video or block of text shows | 2.5s or less | Over 4s |
| Interaction to Next Paint (INP) | How long the page takes to respond to a tap or a key | 200ms or less | Over 500ms |
| Cumulative Layout Shift (CLS) | How far things move about while it loads | 0.1 or less | Over 0.25 |
On most landing pages the LCP is the hero picture, or the headline if there isn't one. A background video can be it too, since a video 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 It replaced First Input Delay in March 2024, so a guide still talking about FID is out of date.
CLS is the one you feel without knowing its name. You go to press "Start free", a picture above it finishes loading (or a cookie banner slides in), and your thumb lands on a link instead. (A move within half a second of your own tap isn't counted, since you caused it.)
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're the people the number is about.
Why your phone score is worse than your laptop's#
Phones are slower, and their connections shakier. The HTTP Archive'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%.
So if you're red on mobile you're in the larger half. That's some comfort. Not much, though, if most of your visitors arrive on their phones from an ad.
How to read a red PageSpeed Insights result#
The top of the PageSpeed Insights page and the bottom are measuring different things, and it'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.
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's changed, so I'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't.
There are two gaps in the field data worth knowing about. A site needs enough visitors to be included at all, and Google doesn'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've got, none of them are in those numbers.
Does speed change how many visitors sign up?#
Probably, though the best evidence is weaker than the figures people quote from it. The study cited most is Deloitte'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%.
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'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't read much more into it than that.
You'll also see "53% of mobile visits are abandoned after 3 seconds" pretty much everywhere. I couldn't find the original 2016 report, and Google's own pages credit the figure to two different studies, so I've left it out.
For what it's worth, I think a slow first screen loses the visitors who were least sure in the first place, and they're exactly who a landing page's first screen is there to win over. That's enough reason for me to fix the causes below without a percentage to point at.
Does page speed affect Google ranking?#
A little, and relevance comes first. Google says "Core Web Vitals are used by our ranking systems" and, in the same documentation, that it "always seeks to show the most relevant content, even if the page experience is sub-par". A good score "doesn't guarantee" a top place either. A fast page that doesn't answer the search will still lose to a slow one that does, which is why your Google title and description usually deserve your time first.
The fixes that matter most on a landing page#
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' scripts. That shape tends to fail in the same few places.
Let the main picture load first#
The hero picture is usually your LCP, so anything that holds it up holds up the number. Web.dev's guide is unusually flat about this, saying "Never lazy-load your LCP image." Lazy loading tells the browser to wait until it knows a picture'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.
The fix is two attributes. Take loading="lazy" off the hero image and add fetchpriority="high", 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.
If the hero is a CSS background, the browser can't find it until the stylesheet has loaded. Preload it from the HTML with <link rel="preload" as="image" fetchpriority="high"> and it'll start straight away.
Give every picture its size#
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's most of a bad CLS, and in 2025 62% of mobile pages had at least one picture with no size set.
Put width and height on every img and video, or give its box a CSS aspect-ratio. Anything that turns up late wants the same treatment (a cookie banner, a review widget, an embedded video), so keep its room with a min-height from the start rather than letting it push the headline down when it lands.
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's size with size-adjust keeps the words where they started.
Move other sites' scripts after the page#
Chat widgets, analytics, heatmaps, A/B testing tools and ad pixels often sit in the page's <head> as plain script tags. A plain script there stops the browser drawing anything until it's downloaded and run, so if one of those companies' servers is slow that morning, your visitor stares at a white screen.
Add defer or async 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'd tell them anyway, and show them the white screen.
Keep taps quick#
A slow INP on a landing page is usually those same scripts, running long jobs on the phone'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'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 "Yielding indiscriminately is better than not yielding at all".
A worked example#
Here's fernly.example, a made-up plant care app, with the phone first screen these pages often have, before and after.
| Before | After |
|---|---|
Hero photo with loading="lazy" and no size | No lazy loading, fetchpriority="high", sized |
Chat widget and heatmap scripts in the head, no defer | Both deferred, the consent tool left as it was |
| Review widget appears above the button once it loads | Its height kept from the start |
| Headline font swaps in wider and reflows to three lines | Font preloaded, fallback matched with size-adjust |
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 its main button staying put where the visitor's thumb expects it.
What Magpie's free check measures#
Magpie's free check loads your site on an iPhone 14 Pro's screen and measures three of these causes. It records the layout shift while the phone's first screen loads, holds it to Google's 0.1 and names what moved. It finds the largest picture on that screen and checks it isn'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' scripts in the head that hold the page up, leaving consent tools out. What it doesn't do is time a single load, since one visit from one place says very little about the slowest quarter of yours.
Questions people ask#
How long until a fix shows in my Core Web Vitals?#
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.
Is it worth paying for a faster host?#
Sometimes. Web.dev's guide to LCP splits it into four parts and, as a rough guide (it calls them "guidelines, not strict rules"), suggests about 40% of the time goes on the server'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'd look at the hero picture and the scripts in the head first, because they cost nothing to fix and I'd bet on them being the problem, and only then start pricing hosts.
Sources
- web.dev, Web Vitals (opens in a new tab)
- web.dev, Largest Contentful Paint (LCP) (opens in a new tab)
- web.dev, Interaction to Next Paint (INP) (opens in a new tab)
- web.dev, Cumulative Layout Shift (CLS) (opens in a new tab)
- web.dev, Optimize Largest Contentful Paint (opens in a new tab)
- web.dev, Optimize Cumulative Layout Shift (opens in a new tab)
- web.dev, Optimize Interaction to Next Paint (opens in a new tab)
- Google Search Central, Understanding page experience in Google Search results (opens in a new tab)
- Google Search Central, Understanding Core Web Vitals and Google search results (opens in a new tab)
- PageSpeed Insights, About PageSpeed Insights (opens in a new tab)
- Chrome UX Report, Methodology (opens in a new tab)
- HTTP Archive, Web Almanac 2025: Performance (Jariyal, Rasam, Humaira and Grogg) (opens in a new tab)
- Deloitte for Google, Milliseconds Make Millions (2020) (opens in a new tab)

About the author
Alex · 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


