Core Web Vitals on mobile: why phone scores lag, and the fixes
Short answer Phone scores trail desktop because phones have slower processors and shakier connections, and PageSpeed's mobile test copies both on purpose. Google ranks the mobile version of your site, so the phone numbers are the ones to fix. Most of the gap comes from oversized top images and heavy scripts like chat widgets and trackers.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callWhy mobile scores are lower than desktop
Two reasons, both physical: a phone has a weaker processor than a laptop, and a cellular connection is slower and less steady than office Wi-Fi. The 2025 Web Almanac blames the mobile loading gap on "the combined effects of slower networks and less capable devices on mobile" (HTTP Archive).
The gap shows up across the whole web, measured on real Chrome visitors in July 2025 (HTTP Archive):
| Share of websites rated "good" | Mobile | Desktop |
|---|---|---|
| All three Core Web Vitals | 48% | 56% |
| LCP (main content shows fast) | 62% | 74% |
| INP (page reacts fast to a tap) | 77% | 97% |
| CLS (layout doesn't jump) | 81% | 72% |
Loading and tap response are where phones lose. Layout jumps are the one place phones do better. So a phone score below your desktop score is normal, not a sign something broke.
What the three metrics and their pass marks mean is covered in Core Web Vitals and Google ranking.
Why PageSpeed's mobile test is so harsh
PageSpeed Insights runs its lab test on a slow phone on purpose. Lighthouse, the engine behind it, applies a "constant 4x CPU multiplier" and limits the connection to 1.6 Mbps download with 150 milliseconds of delay, to copy roughly the 85th-percentile mobile connection (Lighthouse docs). The desktop test uses "an emulated-desktop with a wired connection" (PageSpeed Insights).
Desktop and mobile are also graded on different curves. Google's Lighthouse team says desktop scores "will be significantly different" because each device is scored against its own real-world data (Chrome for Developers). Never compare the two numbers directly.
Your new iPhone on home Wi-Fi feels fast because it is the best case. The test shows the slow case, which is a real share of your visitors.
Which score does Google use: mobile or desktop?
Fix mobile first. Google crawls and ranks the phone version of your pages:
"Google uses the mobile version of a site's content, crawled with the smartphone agent, for indexing and ranking." (Google Search Central)
Speed is a small factor next to relevance. Google says "Core Web Vitals are used by our ranking systems," but it still shows the most relevant page when the experience is weak (Google Search Central). The bigger cost of a slow phone page is the caller who gives up.
How to improve Core Web Vitals on mobile
These five fixes target what phones handle worst. Most can be done on Wix, Squarespace or WordPress without a developer.
| Fix | Metric it helps | Why it matters more on a phone |
|---|---|---|
| 1. Serve a phone-sized top image | LCP | Desktop-sized images on phones "can use 2–4x more data than needed" (web.dev) |
| 2. Load the top image first | LCP | A lazy-loaded hero waits for everything else |
| 3. Cut third-party scripts | INP | A 4x slower processor turns every script into a delay |
| 4. Swap full-screen popups for small banners | Page experience | A popup covers the whole small screen |
| 5. Reserve space for sticky bars and embeds | CLS | A late "Call now" bar pushes the page under a thumb |
1. Serve a phone-sized top image
Your website builder or image plugin should make several sizes of each photo and let the phone pick the small one (the `srcset` attribute does this). Check by opening your top image on a phone and looking at its file size: a 3,000-pixel-wide photo on a 400-pixel-wide screen is a common cause of slow mobile LCP. Saving it as WebP or AVIF shrinks it further, and a content delivery network (CDN) serves it from a server near the visitor (web.dev).
2. Load the top image first
Google's web.dev says: "Never lazy-load your LCP image, as that will always lead to unnecessary resource load delay" (web.dev). Lazy loading is good for photos further down, bad for the first one. The same page suggests marking that one image `fetchpriority="high"`, and only one or two images per page.
3. Cut third-party scripts
Chat widgets, review carousels, heat-map trackers and old ad pixels all run on the phone's processor. Google's INP guide says loading a script "can introduce long tasks on the main thread," which delay the page's response to taps (web.dev). List every add-on, remove what nobody looked at in 90 days, and set the rest to load after the page.
4. Swap full-screen popups for small banners
A coupon popup that covers the page is worse on a phone, where there is nothing else visible. Google's advice:
"Instead of full page interstitials, use banners that take up only a small fraction of the screen" (Google Search Central)
Legally required screens, such as age gates, are exempt (same page).
5. Reserve space for sticky bars and embeds
Phone layouts often add a sticky "Call" or "Book" bar, a map, or a booking widget that loads late. If the page doesn't hold space for it, everything jumps. Give each one a fixed height so it fills a gap instead of pushing content.
How to test your mobile Core Web Vitals
- Open PageSpeed Insights, enter your main service page, and stay on the Mobile tab (PageSpeed Insights).
- Read the real-user section first. It covers the last 28 days of real Chrome visits (same page). The 0–100 lab score is for finding causes.
- Fix the top failing metric, then recheck after 28 days, since real-user data trails.
- Load the page on your own phone with Wi-Fi off and try to call or book.
Example
Illustrative cases, not measured results.
A roofer: desktop 92, mobile 41
The homepage hero is a 5 MB drone photo. Resizing it and turning off lazy loading for that one image moves mobile LCP from about 6 seconds to under 3. Desktop barely changes, since it was never the problem.
A dental office failing INP on phones only
The site runs a chat widget, two review sliders and three tracking pixels. Removing one slider and two unused pixels brings mobile INP under 200 milliseconds over the next month.
A salon with a full-screen coupon popup
Phone visitors close it or leave. The owner replaces it with a slim banner at the bottom, sized so it doesn't push the page.
FAQ
Is a mobile PageSpeed score of 50 bad?
Not by itself. The lab test simulates a slow phone. What counts is whether real visitors pass the three Core Web Vitals.
Why is my mobile score different every time I test?
Changing ads, internet routing, browser extensions and antivirus software all cause swings (Chrome for Developers). Run it several times, or trust the 28-day real-user data.
My site has no mobile real-user data. Now what?
Small sites often have too few Chrome visitors; PageSpeed then shows site-wide data or none (PageSpeed Insights). Use the lab test and your own phone on cellular.
Different for you?
- Want the basics of LCP, INP and CLS and how much they affect ranking: Core Web Vitals and Google ranking.
- Fast on phones, still no calls: website not getting leads.
- Wondering if your builder is holding you back: Wix, Squarespace or WordPress for SEO.
- Building a trade site meant for phone callers: HVAC and plumber website design.
- Want the site made fast and kept fast: Website Agent.