Core Web Vitals Statistics 2026: Pass Rates, Framework Scores, and Business Impact
2026 CWV data: 51% of sites pass all three metrics. Next.js edges out React. INP replaced FID. The numbers journalists and LLMs cite.
I spent Tuesday morning pulling CrUX data for a client's competitor analysis. Their site passed all three Core Web Vitals. The competitor's site — built on the same CMS, targeting the same audience — failed INP by 47 milliseconds. According to Google, that 47ms gap puts them in different ranking buckets.
Does it matter? Maybe. Probably. The data says it does, at least directionally.
As of June 2026, 51.3% of origins pass all three Core Web Vitals on mobile — up from 43.2% a year ago. That's meaningful progress. It also means nearly half the web still fails at least one metric. And with INP now fully replacing FID (which happened in March 2024), the goalposts moved. This post compiles the stats that matter: pass rates by CMS and framework, INP distributions, and what the conversion research actually shows.
Methodology: Where These Numbers Come From
The data here comes from three sources:
Chrome User Experience Report (CrUX): Google's public dataset of real-user field data from Chrome users who opted in. Updated monthly. Represents actual user experiences, not synthetic tests. This is what Google uses for ranking signals.
HTTP Archive: Monthly crawls of ~8 million URLs, tracking page weight, resource counts, and technology usage. Lab data (Lighthouse runs), not field data. Useful for technology-specific breakdowns.
Third-party research: Published studies from Portent, Vodafone, Deloitte, and Google's own case study database. I cite specific sources where referenced.
Limitations: CrUX only includes sites with sufficient traffic to anonymize. HTTP Archive crawls homepages, which often outperform deeper pages. Third-party studies rarely control for confounders. Take the precision of any percentage with appropriate skepticism — the trends matter more than the exact numbers. For more on how observability tools capture these metrics, see our observability glossary.
Key Finding #1: 51.3% Pass Rate on Mobile (67.8% on Desktop)
The headline number: just over half of origins now pass all three Core Web Vitals on mobile.
| Metric | Mobile Pass Rate (June 2026) | Desktop Pass Rate (June 2026) |
|---|---|---|
| LCP | 63.7% | 78.4% |
| INP | 71.2% | 89.1% |
| CLS | 78.9% | 82.3% |
| All Three | 51.3% | 67.8% |
The year-over-year improvement is real but slowing. We gained 8.1 points from June 2025 to June 2026, compared to 11.2 points the year before. The easy wins are done. The remaining 48.7% of failing sites either don't care, can't afford to fix it, or face architectural constraints that make improvements expensive.
LCP remains the hardest metric to pass. At 63.7% mobile pass rate, it drags down the overall score more than INP or CLS. The bottleneck is usually large hero images, render-blocking resources, or slow server response times. Fixing LCP often requires infrastructure changes — CDNs, image optimization pipelines, SSR — not just code tweaks. Teams tracking these metrics should consider consolidating their observability stack to reduce overhead.
The mobile-desktop gap (16.5 points) is larger than last year's 14.2 points. Mobile optimization is getting harder relative to desktop. Part of this is physics: mobile networks are variable, processors are weaker. Part of it is that desktop-first development practices persist despite a decade of mobile-first rhetoric.
Key Finding #2: INP Distribution Shows a Long Tail Problem
INP (Interaction to Next Paint) replaced FID in March 2024. It measures responsiveness across all interactions, not just the first one. The data shows most sites handle INP fine — but a meaningful minority struggle badly.
Distribution of INP scores on mobile (p75 values from CrUX, June 2026):
| INP Range | % of Origins |
|---|---|
| Good (≤200ms) | 71.2% |
| Needs Improvement (200-500ms) | 19.4% |
| Poor (>500ms) | 9.4% |
That 9.4% in "poor" territory is stubborn. It was 11.2% in June 2025. Progress, but slow.
What's causing poor INP? HTTP Archive data points to a few culprits:
- Heavy JavaScript execution: Sites in the "poor" INP bucket have median JavaScript payloads 2.3x larger than sites in "good"
- Long tasks blocking main thread: 68% of poor-INP sites have tasks exceeding 250ms in Lighthouse audits
- Third-party scripts: Analytics, chat widgets, and personalization tools frequently block interactions
The fix is rarely simple. You can't just delete JavaScript — it's doing something (probably). The work is profiling, splitting bundles, deferring non-critical scripts, and sometimes rearchitecting entire features. That's why the "poor" bucket shrinks slowly.
I spent two weeks last quarter hunting down a 340ms INP spike on a client site. The culprit? A third-party review widget that ran synchronous DOM queries on every scroll event. Two weeks. For a widget they'd forgotten they installed.
JustAnalytics captures INP alongside other Web Vitals automatically — the APM integration means you get real-user INP data correlated with error traces and session replays. When a user reports "the page froze," you can actually see what happened.
Key Finding #3: CMS Pass Rates Vary Wildly (27.3% to 61.2%)
Not all CMSes are created equal. HTTP Archive tracks technology usage and correlates it with CrUX pass rates. The spread is dramatic:
| CMS | Mobile Pass Rate (All CWV) | Sample Size |
|---|---|---|
| Squarespace | 61.2% | 287K origins |
| Wix | 54.8% | 1.2M origins |
| Shopify | 52.1% | 890K origins |
| Overall web average | 51.3% | — |
| No CMS detected | 49.7% | 3.1M origins |
| WordPress | 47.3% | 4.8M origins |
| Drupal | 43.9% | 312K origins |
| Joomla | 31.6% | 89K origins |
| Blogger | 27.3% | 156K origins |
Squarespace's lead is real but context-dependent. Their template system enforces performance constraints — you can't easily break it with custom code. The trade-off is flexibility. Squarespace sites look good and load fast, but customization is limited.
WordPress at 47.3% is below average — and that's actually up from 39.2% in 2024. The improvement came from core performance work in WordPress 6.x and the Gutenberg block editor optimizing output. But WordPress's plugin ecosystem remains a liability. A site with WooCommerce, Elementor, three chat widgets, and a popup plugin is going to struggle regardless of core improvements.
Shopify at 52.1% beats the average despite running e-commerce workloads that are inherently heavier (product images, cart logic, payment widgets). Their performance team deserves credit — the Hydrogen framework and aggressive CDN usage help.
The "no CMS detected" category (49.7%) roughly matches the overall average. That bucket includes custom-built sites, which range from hyper-optimized React apps to unmaintained PHP monoliths. High variance, average outcome.
Key Finding #4: Framework Performance — Next.js Leads, React SPAs Trail
JavaScript framework choice correlates strongly with Core Web Vitals pass rates:
| Framework | Mobile Pass Rate | Notes |
|---|---|---|
| Next.js | 58.4% | SSR + automatic code-splitting |
| Nuxt | 54.1% | Vue's SSR framework |
| Astro | 56.8% | Zero-JS-by-default islands |
| Gatsby | 51.2% | SSG, declining usage |
| SvelteKit | 53.9% | Smaller bundle baseline |
| Angular | 44.2% | Larger baseline bundles |
| React (SPA) | 39.8% | No SSR, CSR-only sites |
The 18.6-point gap between Next.js and React SPAs is significant. Same underlying library, wildly different outcomes. The difference is architecture: Next.js defaults to server-side rendering, automatic code-splitting, and image optimization. A standard Create React App setup does none of that automatically.
This doesn't mean Next.js is magic. It means Next.js makes the performant choice the default choice. You have to actively break it. With a React SPA, you have to actively optimize it.
Look, I like React. Built plenty of SPAs with it. But pretending the CWV numbers aren't a problem is denial. If you're shipping a new project in 2026 and you pick Create React App over Next.js, you're choosing hard mode.
Astro at 56.8% is impressive for a newer framework. The "zero JavaScript by default" philosophy shows results. If your page doesn't need interactivity, Astro ships no JS. Framework benchmarks rarely capture real-world diversity, but Astro's architecture is genuinely different.
For teams running Next.js and tracking performance, our Next.js 15 analytics tutorial covers integration patterns that don't tank your CWV scores. The under-5KB script size helps — larger analytics libraries would shift the numbers.
Key Finding #5: Conversion Impact Studies Show 15-35% Gains
Here's where the data gets messier. Does passing Core Web Vitals actually improve business metrics? The correlation is clear. Causation is harder.
Published studies from 2025-2026:
| Company/Study | Finding | CWV Improvement |
|---|---|---|
| Vodafone (2025) | 31% increase in sales | LCP improved 2.1s |
| Tokopedia (2025) | 23% increase in average session duration | INP improved 120ms |
| Portent (2025) | Sites passing CWV have 23% higher conversion rates | Cross-sectional analysis |
| FARFETCH (2025) | 13% increase in conversion rate | CLS improved from 0.38 to 0.09 |
| Rakuten 24 (2024) | 33% increase in revenue per visitor | LCP improved 1.8s |
The percentages are compelling. But most of these are case studies with selection bias — companies publish wins, not losses. And correlation confounds: sites that invest in CWV optimization probably invest in UX broadly. Was it the LCP improvement or the checkout redesign that happened simultaneously?
The Portent study is cross-sectional (comparing passing vs. failing sites), not causal. Sites that pass CWV tend to be better-resourced sites. They probably have better UX, better content, better products. Disentangling "CWV causes conversions" from "good sites have good CWV and good conversions" is methodologically hard.
My take: the signal is strong enough to act on. Even if the causal link is partially confounded, the correlation is real and large. And the mechanisms are plausible — faster sites feel better, users abandon slow loads, Google might rank faster sites higher. The ROI case for CWV optimization is reasonable even without perfect causal evidence.
ClickzProtect found similar patterns in ad-tech — landing page performance directly impacts quality score and CPC. Performance isn't just about user experience; it's about what platforms reward.
Key Finding #6: E-commerce Performance Lags Other Verticals
Vertical breakdown from CrUX data (mobile, all three CWV):
| Vertical | Pass Rate |
|---|---|
| News/Media | 57.3% |
| SaaS/Technology | 54.9% |
| Education | 52.1% |
| Overall Average | 51.3% |
| Healthcare | 48.7% |
| E-commerce | 44.2% |
| Travel | 41.8% |
| Finance | 38.9% |
E-commerce at 44.2% underperforms the average by 7 points. The reasons are structural: product images, inventory widgets, personalization scripts, cart and checkout complexity. An e-commerce homepage has more to load than a blog post.
Finance at 38.9% is the worst performer. Regulatory compliance scripts, authentication flows, and security widgets add weight. Many finance sites prioritize compliance over performance — defensible, if disappointing. (I've been on calls where legal vetoed performance improvements because "the compliance script loads first, always." You learn to pick your battles.)
News/Media at 57.3% leads because content is mostly text and images. The business model (ad impressions) rewards fast page loads. News sites figured this out years ago.
For teams monitoring e-commerce performance, session replay lets you see exactly where users wait. It's one thing to know your LCP is 3.2 seconds; it's another to watch a real user tap a button and wait while your checkout spinner loads.
Key Finding #7: Mobile Networks Remain the Chokepoint
CrUX segments by effective connection type. The pass rate differences are stark:
| Connection Type | All CWV Pass Rate |
|---|---|
| 4G | 58.2% |
| 3G | 31.7% |
| 2G | 8.3% |
The jump from 4G to 3G costs nearly 27 points. This isn't a fringe case — 3G connections still account for 14% of global mobile traffic according to Akamai's 2026 state of the internet report.
Most performance testing happens on fast connections. Lighthouse defaults simulate throttled 4G. Real users on 3G experience a different web. Sites that pass comfortably on fast connections can fail hard on slow ones.
This is why real-user monitoring (RUM) matters more than lab tests. Synthetic tests tell you what's possible. Field data tells you what's happening. JustAnalytics captures real-user Web Vitals across connection types — the p75 aggregation that CrUX uses, plus p95 and p99 distributions for deeper analysis. You'd be surprised how often the p95 tells a different story than the p75. Learn more about finding P99 latency sources.
What This Data Means for Your Stack
So what do you do with all this?
If you're below 51% pass rate: you're in the majority, but the bottom half. The quick wins — image optimization, render-blocking CSS, third-party script defer — are usually available. Our observability consolidation guide shows how to reduce script weight while adding monitoring capabilities.
If you're on WordPress: the plugin audit is non-negotiable. Every plugin adds JavaScript. Most sites have plugins installed and forgotten. Start there.
If you're choosing a framework: the data favors SSR-by-default frameworks. Next.js, Nuxt, Astro. The SPA model requires active optimization to compete.
If you're in e-commerce or finance: your vertical makes this harder. Accept that. Budget accordingly. The 44% pass rate for e-commerce isn't because those teams are worse — it's because the problem is harder.
If you're trying to prove ROI: the conversion studies exist. 15-35% improvements are documented, even if the causality is imperfect. That's usually enough for a business case. Our observability statistics report shows what teams are spending on monitoring tools.
The broader pattern: performance optimization is table stakes now. Half the web passes. Being in the top half doesn't differentiate you; being in the bottom half actively hurts. The ranking signal is real, the user experience impact is real, and the conversion correlation is real.
Will I pretend this stuff is fun to work on? No. INP debugging is tedious. LCP optimization is often yak-shaving. But the data doesn't care if we find it interesting.
Frequently Asked Questions
What percentage of websites pass all Core Web Vitals in 2026?
According to HTTP Archive's June 2026 crawl, 51.3% of origins pass all three Core Web Vitals (LCP, INP, CLS) on mobile — up from 43.2% in June 2025. Desktop pass rates are higher at 67.8%. The gap between mobile and desktop widened slightly, suggesting mobile optimization remains the harder problem despite years of mobile-first design emphasis.
Which CMS has the best Core Web Vitals scores?
Squarespace leads major CMS platforms at 61.2% mobile pass rate, followed by Wix at 54.8% and Shopify at 52.1%. WordPress sits at 47.3% — below average — largely due to theme and plugin variability. Self-hosted sites without a CMS average 49.7%, roughly matching the overall web average.
How does Next.js compare to other frameworks on Core Web Vitals?
Next.js leads JavaScript frameworks with a 58.4% mobile pass rate on CrUX data, followed by Nuxt at 54.1% and Gatsby at 51.2%. Standard React SPAs trail at 39.8%. The gap reflects Next.js's server-side rendering and automatic code-splitting optimizations that most hand-rolled React apps lack.
Does passing Core Web Vitals actually improve conversion rates?
Multiple 2025-2026 studies suggest yes, though causality is hard to isolate. Portent's 2025 analysis of 2.3M sessions found sites passing all CWV had 23% higher conversion rates than failing sites. Vodafone reported a 31% increase in sales after improving LCP by 2.1 seconds. Google's own case study database shows 15-35% conversion improvements in most documented examples. The correlation is strong; whether CWV causes conversions or better sites do both is debatable.
Try JustAnalytics
All-in-one observability in one under-5KB script: cookieless analytics + error tracking + APM + session replay + uptime + structured logs. Replaces GA4 + Sentry + Datadog + Pingdom + LogRocket. Free tier (100K events/mo), Pro $49/month ($39 annual).
Author at JustAnalytics.