Only 48% of mobile websites pass Core Web Vitals, and the single biggest reason is images: they account for 76% of Largest Contentful Paint elements on mobile, yet 57% of them are still JPG. Desktop fares better at 56% passing, but it has barely moved in a year while mobile climbed from 36% in 2023 to 48%. The metric that looks healthiest in aggregate hides the sharpest inversion in the dataset: Interaction to Next Paint is 97% good on desktop but only 63% good among the top 1,000 mobile sites. The figures below come from the HTTP Archive Web Almanac and Chrome UX Report field data.
TL;DR
- 48% of mobile sites have good Core Web Vitals (HTTP Archive Web Almanac)
- 56% of desktop sites have good Core Web Vitals (HTTP Archive Web Almanac)
- Mobile improved from 36% in 2023 to 44% in 2024 to 48% (HTTP Archive)
- Desktop moved just one point, from 55% (HTTP Archive)
- LCP is good on 62% of mobile pages and 74% of desktop pages (HTTP Archive)
- Images are 76% of mobile LCP elements and 85.3% of desktop (HTTP Archive)
- JPG is still 57% of LCP images, down 4 points (HTTP Archive)
- WebP reached 11%, up 4 points, and AVIF just 0.7% (HTTP Archive)
- INP is good on 97% of desktop and 77% of mobile pages (HTTP Archive)
- Only 63% of the top 1,000 mobile sites are good on INP (HTTP Archive)
- CLS is good on 81% of mobile pages, the one metric where mobile leads (HTTP Archive)
- 65% of desktop pages have images without explicit dimensions (HTTP Archive)
- 87% of mobile pages use web fonts (HTTP Archive)
1. The Headline Pass Rates
Two numbers describe the state of web performance and they are moving at different speeds. 48% of mobile sites and 56% of desktop sites have good Core Web Vitals, with mobile up 4 points year over year and desktop up just 1. Mobile is closing the gap not because desktop got worse but because desktop stopped improving.
| Metric | Value | Source |
|---|---|---|
| Good Core Web Vitals, mobile | 48% | HTTP Archive Web Almanac |
| Good Core Web Vitals, desktop | 56% | HTTP Archive Web Almanac |
| Mobile figure in 2024 | 44% | HTTP Archive |
| Mobile figure in 2023 | 36% | HTTP Archive |
| Desktop figure in 2024 | 55% | HTTP Archive |
| Mobile improvement year over year | 4 points | HTTP Archive |
| Desktop improvement year over year | 1 point | HTTP Archive |
| Gap between desktop and mobile | 8 points | Derived from HTTP Archive figures |
| Data collection | July 2025, CrUX field data | HTTP Archive |
Mobile has gained 12 points in two years against desktop’s near-stall, which suggests the remaining desktop failures are structural rather than the kind of low-hanging fruit mobile teams have been picking. Source: HTTP Archive Web Almanac performance chapter.
2. Largest Contentful Paint Is an Image Problem
LCP is the metric that fails most, and its cause is unusually well identified. Images account for 85.3% of LCP elements on desktop and 76% on mobile, while only 62% of mobile pages achieve a good LCP score. Text-based LCP is a minority case at 14.4% desktop and 23.7% mobile, which means most LCP work is image delivery work rather than rendering work.
| Metric | Value | Source |
|---|---|---|
| Good LCP, desktop | 74% | HTTP Archive |
| Needs improvement, desktop | 18% | HTTP Archive |
| Poor LCP, desktop | 7% | HTTP Archive |
| Good LCP, mobile | 62% | HTTP Archive |
| Needs improvement, mobile | 25% | HTTP Archive |
| Poor LCP, mobile | 13% | HTTP Archive |
| Images as LCP element, desktop | 85.3% | HTTP Archive |
| Images as LCP element, mobile | 76% | HTTP Archive |
| Text as LCP element, mobile | 23.7% | HTTP Archive |
| Pages lazy-loading their LCP image | 16% to 17% | HTTP Archive |
That last row is the self-inflicted wound of the dataset: roughly one page in six lazy-loads the very image that determines its LCP score, which delays the metric it most needs to improve. Source: HTTP Archive Web Almanac performance chapter.
3. Image Formats Are Moving Slowly
Given that images drive LCP, format adoption is the leading indicator worth watching, and it is unhurried. JPG remains 57% of LCP images, down 4 points, while WebP reached 11%, up 4, and AVIF sits at 0.7%. At roughly 4 points of migration per year, JPG stays the majority LCP format well into the 2030s.
| Metric | Value | Source |
|---|---|---|
| JPG share of LCP images | 57% | HTTP Archive |
| Change year over year | down 4 points | HTTP Archive |
| PNG share | 26% | HTTP Archive |
| WebP share | 11% | HTTP Archive |
| WebP change year over year | up 4 points | HTTP Archive |
| AVIF share | 0.7% | HTTP Archive |
| fetchpriority=“high” adoption, mobile | 17.3% | HTTP Archive |
| fetchpriority=“high” adoption, desktop | 16.3% | HTTP Archive |
| Preload adoption | 2.1% to 2.2% | HTTP Archive |
The fetchpriority numbers are the more actionable pair: a hint that directly addresses LCP is used on roughly one page in six, and preload on one in fifty. Accessibility-side data sits in our web accessibility statistics. Source: HTTP Archive Web Almanac performance chapter.
4. Responsiveness Inverts With Popularity
Interaction to Next Paint looks like the solved metric until you segment it. INP is good on 97% of desktop pages and 77% of mobile pages, but only 63% of the top 1,000 mobile sites, which is 14 points below the overall mobile average. The most-visited sites are the least responsive, because they carry the most JavaScript.
| Metric | Value | Source |
|---|---|---|
| Good INP, desktop | 97% | HTTP Archive |
| Good INP, mobile | 77% | HTTP Archive |
| Mobile change year over year | up 3 points | HTTP Archive |
| Good INP, top 1,000 mobile sites | 63% | HTTP Archive |
| Same figure a year earlier | 53% | HTTP Archive |
| Improvement in the top 1,000 | 10 points | HTTP Archive |
| Good INP, top 10,000 mobile sites | 56% | HTTP Archive |
| Good INP, top 100,000 mobile sites | 56% | HTTP Archive |
| Gap between top 1,000 and overall mobile | 14 points | Derived from HTTP Archive figures |
The top 1,000 improved 10 points in a year, faster than any other cohort, which suggests the largest sites have both the worst problem and the most resources to attack it. The flat reading of 56% across both the top 10,000 and the top 100,000 is the other half of that story: outside the very largest properties, responsiveness does not vary with popularity at all, so the JavaScript burden that degrades INP is close to uniform across the commercial web. Ranking context sits in our SEO statistics and rendering-engine context in our browser market share statistics. Source: HTTP Archive Web Almanac performance chapter.
5. Layout Shift and the Unsized Image Epidemic
CLS is the one metric where mobile beats desktop, and its dominant cause is a one-line fix that most pages have not applied. 81% of mobile pages and 72% of desktop pages have good CLS, yet 65% of desktop pages and 62% of mobile pages contain images with no explicit dimensions. The median page has 2 such images; the 90th percentile has 26 on desktop.
| Metric | Value | Source |
|---|---|---|
| Good CLS, mobile | 81% | HTTP Archive |
| Good CLS, desktop | 72% | HTTP Archive |
| Desktop pages with unsized images | 65% | HTTP Archive |
| Mobile pages with unsized images | 62% | HTTP Archive |
| Median unsized images per page | 2 | HTTP Archive |
| 90th percentile, desktop | 26 | HTTP Archive |
| 90th percentile, mobile | 23 | HTTP Archive |
| Mobile pages using web fonts | 87% | HTTP Archive |
| Pages with non-composited animations, desktop | 44% | HTTP Archive |
| Pages with non-composited animations, mobile | 40% | HTTP Archive |
| Early Hints adoption, desktop top 1M | barely 6% | HTTP Archive |
| Speculation Rules adoption, top 10M | 24% to 25% | HTTP Archive |
That 62% of mobile pages carry unsized images while 81% still pass CLS tells you most unsized images sit below the fold where they shift nothing. The risk is latent rather than realised, and it materialises the moment layout changes, a font loads slowly, or a slower connection reorders the paint sequence. Connection-quality context sits in our internet speed statistics. Source: The 2025 Web Almanac.
Summary: Website Performance by the Numbers
| Metric | Value | Source |
|---|---|---|
| Good Core Web Vitals, mobile | 48% | HTTP Archive |
| Good Core Web Vitals, desktop | 56% | HTTP Archive |
| Mobile CWV in 2023 | 36% | HTTP Archive |
| Mobile CWV in 2024 | 44% | HTTP Archive |
| Good LCP, mobile | 62% | HTTP Archive |
| Good LCP, desktop | 74% | HTTP Archive |
| Poor LCP, mobile | 13% | HTTP Archive |
| Images as LCP element, desktop | 85.3% | HTTP Archive |
| Images as LCP element, mobile | 76% | HTTP Archive |
| Pages lazy-loading the LCP image | 16% to 17% | HTTP Archive |
| JPG share of LCP images | 57% | HTTP Archive |
| WebP share | 11% | HTTP Archive |
| AVIF share | 0.7% | HTTP Archive |
| fetchpriority=“high” adoption, mobile | 17.3% | HTTP Archive |
| Good INP, desktop | 97% | HTTP Archive |
| Good INP, mobile | 77% | HTTP Archive |
| Good INP, top 1,000 mobile sites | 63% | HTTP Archive |
| Good CLS, mobile | 81% | HTTP Archive |
| Good CLS, desktop | 72% | HTTP Archive |
| Desktop pages with unsized images | 65% | HTTP Archive |
| 90th percentile unsized images, desktop | 26 | HTTP Archive |
| Mobile pages using web fonts | 87% | HTTP Archive |
| Good FCP, mobile | 55% | HTTP Archive |
| Good TTFB, mobile | 44% | HTTP Archive |
Methodology and Sources
- All Core Web Vitals pass rates, metric breakdowns, LCP element composition, image format shares, resource hint adoption, layout shift data, and cohort comparisons come from the HTTP Archive Web Almanac performance chapter, based on July 2025 measurements combining the HTTP Archive crawl with Chrome UX Report field data from real Chrome users (performance chapter, full Web Almanac, page weight chapter, SEO chapter, HTTP Archive).
- Historical context on the metrics themselves comes from published Core Web Vitals history (Addy Osmani), and independent summaries of the 2025 edition were checked against the primary chapters (LavX analysis).
- Data watch: the most recent full Web Almanac edition draws on July 2025 measurements, so these figures are roughly a year old at time of writing and are flagged as most recent available at this level of detail. Chrome UX Report data itself updates monthly and current month-level pass rates will differ from the annual analysis. CrUX covers Chrome users who opt into usage reporting, so it excludes other browsers entirely and underrepresents regions and devices where Chrome share is lower. Pass rates are computed at the origin level using 75th-percentile thresholds, which means a site can fail while most of its visits are fast. Rows marked as derived are arithmetic on published figures.
- Last updated: August 2, 2026. We update this roundup quarterly, and the next major refresh is expected when HTTP Archive publishes its 2026 Web Almanac.