Two-Thirds of the Sites That Fail Core Web Vitals Fail on Just One Metric

Failing Core Web Vitals sounds like a verdict on the whole site: slow, janky, broken across the board. For most sites, it is nothing of the kind. Two-thirds of failing sites fall short on just one of the three metrics, which means most of the failing web is not broken. It is a single fix away from passing.

That distinction is worth money. Site owners who assume a failing grade means a ground-up rebuild often do nothing at all, put off by the scale of the job, while their pages keep losing visitors to slow loads and shedding search rankings. Core Web Vitals have been a Google ranking signal since 2021, and a page that fails is paying for it in conversions, trust, and traffic. Knowing that the fix is usually one targeted change turns an overwhelming problem into an afternoon's work.

We analyzed how sites actually fail, using Google's Chrome User Experience Report, the real-user field data collected from Chrome visitors across nearly 14 million websites. The pattern is consistent and encouraging: failure is concentrated, not diffuse, and on mobile the one metric to check is almost always the same one.

The figures come from the Chrome User Experience Report, Google's record of real-user experience, restricted to the 13.7 million origins (8.17M mobile, 5.59M desktop) that have all three Core Web Vitals present. Each origin is assessed at the 75th percentile against Google's thresholds: loading (LCP) good at 2.5 seconds or less, responsiveness (INP) good at 200 ms or less, and visual stability (CLS) good at 0.10 or less. A site passes only when all three are good at once, and we classified every failing site by how many of the three metrics it misses.

Core Web Vitals is pass or fail, and a site passes only if it is good on all three metrics, which makes failing sound worse than it usually is. Among sites with a full set of data, 30.1% on mobile fail on exactly one metric, while only 15.2% fail on two or three. Single-metric failures outnumber multi-metric failures two to one. Of all the mobile sites that fail, roughly two in three are good on two of the three vitals and fall short on just one.

And that one fix is usually the same one. On mobile, loading (LCP) is the lone culprit far more often than anything else: 18.8% of all sites fail on LCP alone, more than INP-only (6.8%) and CLS-only (4.5%) put together. That 6.8% INP-only figure is measured among sites with all three Core Web Vitals present; expressed as a share of the entire mobile web (including sites without a full reading), the same group is about 5.0%, the figure reported in the INP device-gap study. If you picked a random failing mobile site, the single most likely situation is that its loading is red and its responsiveness and stability are green.

That makes the web's aggregate Core Web Vitals problem far more tractable than it looks. It is concentrated, not diffuse, and it points to a clear first move. For most failing sites, the work is not a rebuild. It is finding the one red metric, which is probably loading, and fixing that.

The mechanics behind the mobile pattern explain why loading dominates. Phones combine slower networks, weaker processors, and pages that were often designed for desktop first, and all three pressures land hardest on the moment the main content has to appear. Responsiveness and stability, by contrast, are governed by choices that modern platforms increasingly handle by default, such as reserving space for images and shipping lighter scripts. So when a mobile site fails, the odds favor loading being the lone holdout, and the fix is a known set of moves rather than a mystery.

The two devices point to different single fixes. On mobile, LCP dominates the one-metric failures, with responsiveness (INP) a meaningful secondary at 6.8%. The mobile web's fixable problems are mostly loading, then responsiveness. On desktop, the picture flips toward stability: loading (14.4%) and CLS (12.4%) are neck and neck as lone culprits, and INP is almost never the sole problem (0.6%), because desktop hardware keeps responsiveness green almost everywhere.

Desktop failures are also shallower. Only 7.0% of desktop sites fail on two or more metrics, versus 15.2% on mobile, so a failing desktop site is very likely one fix from passing. A failing mobile site is more likely to have compounding problems, though even there, single-metric failures are the clear majority.

Among origins with all three metrics measured (8.17M mobile, 5.59M desktop), May 2026 CrUX. Because this pass rate counts only sites with all three Core Web Vitals present, it runs above the stricter whole-web figure of about 40% that treats sites without an INP reading as failing.

If your site fails Core Web Vitals, do not assume a rebuild, because two-thirds of failing sites fall short on a single metric. Open your field data and find the one that is red. On mobile it is most likely loading, which fails alone on 18.8% of all sites, more than INP-only (6.8%) and CLS-only (4.5%) put together. Optimize and prioritize your largest image and defer the render-blocking resources in front of it, and that one lever moves the most sites across the line.

If loading is already green, your one thing is probably responsiveness on mobile, where INP fails alone on 6.8% of sites, or stability on desktop, where CLS does at 12.4%. Desktop failures are shallower still, with only 7.0% of sites failing on two or more metrics against 15.2% on mobile, so a failing desktop site is very likely a single change from the 65.6% that already pass. For most of the web, a red grade is a targeted problem with a targeted fix.