Core Web Vitals Benchmarks 2026: What Good Looks Like

Core Web Vitals benchmarks only matter as a reference point — a way to answer “is my LCP good?” without guessing. As of May 2026 Chrome UX Report (CrUX) data, only 55.9% of all tracked origins globally pass all three Core Web Vitals. That single number is the baseline every site should be measured against before anyone argues about optimization tactics.

This is a benchmark reference, not a how-to. If you already know your scores are weak and want the fixes, we cover those in the Core Web Vitals optimization guide. Here, the job is to set the bar: what the thresholds are, how far mobile lags desktop, which metric is hardest to pass, and how much your platform choice moves the odds before you write a line of code.

Every figure below is sourced and dated. Where the underlying data is vendor-reported, single-source, or unconfirmed, we say so rather than dress a soft number up as a hard one. CrUX field data is the spine; the 2025 Web Almanac supplies the device and year-over-year breakdowns; web.dev supplies the official thresholds.

Start with the one figure worth memorizing: as of the May 2026 CrUX release (published June 9, 2026), 55.9% of all tracked origins pass all three Core Web Vitals. CrUX measured 18,445,974 origins that month. The individual metric pass rates that month were 68.6% good LCP, 81.3% good CLS, and 86.6% good INP. These are vendor-reported figures from Google’s own release notes — directionally solid, but worth re-checking against the dashboard for your own origin.

What makes that 55.9% the right yardstick is the rule behind it. A page only earns an overall good Core Web Vitals assessment when 75% of real user visits individually clear each threshold — not the median visit, the 75th percentile. And all three metrics must pass at once. A site with flawless INP and CLS but a poor LCP still fails the overall assessment. That is why a respectable-looking dashboard can still read as failing: one weak metric sinks the whole grade.

The most recent data also carries a caveat the benchmark itself should surface. The 55.9% is down 0.8 percentage points from April 2026 — a small but real regression. Google has publicly noted it suspects an Android-specific cause but has been explicit that nothing is definitive. Read the dip as a reminder that CrUX is a moving target, not as a confirmed structural shift.

Before benchmarking against anyone, fix the thresholds in your head. They have not changed for LCP and CLS, and INP’s have held since it became a Core Web Vital in March 2024. One myth worth killing first: the LCP good threshold is 2,500 ms (2.5 seconds), not 2.0 seconds. A stray claim that it dropped to 2.0 s in 2026 is wrong — web.dev still documents 2.5 s as the unchanged bar.

INP replaced First Input Delay (FID) as a Core Web Vital on March 12, 2024. The distinction matters when reading older benchmarks: FID only measured the delay of the first interaction, while INP captures responsiveness across every interaction in a session — a far harder bar to clear, and the reason responsiveness numbers shifted when the metric changed. Any pass-rate data from before that date is measuring a different, easier thing.

The single most important context for any benchmark is which device it describes. Per the 2025 Web Almanac (CrUX data from July 2025), 48% of mobile origins pass all three Core Web Vitals versus 56% of desktop origins — an 8 percentage-point gap. If you only ever look at a desktop score, you are flattering yourself by roughly eight points against the surface most of your users are actually on.

Two patterns sit inside that gap. First, the trend is genuinely improving on mobile: pass rates climbed from 32% (2021) to 36% (2023) to 48% (2025), while desktop rose from 41% to 56% over the same span. The web is getting faster, just unevenly. Second, secondary pages consistently beat home pages — 61% versus 47% on desktop, 56% versus 45% on mobile — because home pages tend to carry the heaviest hero media and the largest LCP element. When you benchmark, compare like with like: a home page against home pages, an article against articles.

Not all three metrics are equally hard to pass, and benchmarking without that nuance leads teams to optimize the wrong thing. On mobile (July 2025 CrUX), only about 62% of pages achieve a good LCP, against roughly 77% for INP and 81% for CLS. Loading is the bottleneck. If your site is failing the overall assessment, LCP is the statistically most likely culprit — and the place to look first.

INP is worth a second look because of where it fails. Desktop INP passes at roughly 96.8%; mobile is markedly lower. One vendor data study put median p75 INP at about 120 ms on desktop versus about 248 ms on mobile — directional figures we have not independently confirmed against the CrUX API, so treat them as illustrative of the gap rather than precise. The mechanism is uncontroversial even if the exact numbers are: weaker mobile processors and heavier JavaScript execution make interactions slower to acknowledge on phones than on laptops.

Before a single optimization, your platform has already placed a large bet on your behalf. On November 2025 HTTP Archive data, Duda passed at 84.87% while WordPress sat at 46.28% — a 38-point spread driven mostly by defaults: theme weight, image handling, and render-blocking scripts. This is the most actionable benchmark in the whole reference, because platform choice is a decision you make once and live with.

The platform numbers also explain a real pattern we see in audits. Two sites on the same CMS can read very differently because the platform sets the floor, not the ceiling — a well-built WordPress site with a lightweight theme and disciplined image handling can pass comfortably, while a Duda or Shopify site stuffed with third-party scripts can fail. The platform tilts the odds; your build decides the outcome. If you are choosing a stack from scratch, weighting Core Web Vitals into that decision is exactly the kind of judgment our web development engagements bring to the table early, before the defaults lock in.

The most expensive benchmarking mistake is comparing the wrong number. Google assesses Core Web Vitals for ranking from CrUX field data — real Chrome users, a 28-day rolling window, the 75th percentile — not from Lighthouse lab scores. Lighthouse runs a single simulated load on one machine; it is a diagnostic, not the thing Google ranks on. A perfect Lighthouse 100 and a failing CrUX assessment can coexist on the same URL.

Yes — but the honest answer needs hedging, and a benchmark reference owes you the hedge. Core Web Vitals is one ranking signal among many, and Google has never published a quantified weight for it. The clearest framing comes from Google Search Advocate John Mueller, who has described it as more consequential than a pure tie-breaker while being explicit that it does not override relevance.

Google’s own documentation reinforces the same boundary: there is no single signal, and the most relevant content can still rank despite a sub-par page experience. Where many pages offer comparably helpful content, a great page experience can be the edge. That is the realistic mental model — Core Web Vitals as a contributor and a differentiator in close races, not a lever that overrides weak content or thin relevance.

You will see a widely cited correlation that pages ranking first are roughly 10% more likely to pass Core Web Vitals than pages at position nine. Read it as correlation, not causation: both a high ranking and a passing CWV grade tend to reflect overall site quality and investment, so they travel together without one cleanly causing the other. Treating that 10% as a causal dial to turn is exactly the kind of over-reading the data does not support.

The business case for passing does not actually rest on the ranking question. Google-Deloitte joint research is widely cited for the finding that a 0.1-second speed improvement can lift retail conversion rates and average order value — and web.dev’s case studies document real outcomes: Vodafone Italy improved LCP and reported 8% more sales; Rakuten 24 optimized all three metrics and saw materially higher revenue per visitor and conversion. The exact Google-Deloitte percentages sit behind a paywalled study, so treat the specific figures as indicative rather than gospel; the direction — faster pages convert better — is well evidenced.

Benchmarks are only useful if you can place yourself on them. Pull your origin’s field data from PageSpeed Insights or the CrUX dashboard, then read it against the bars below — overall first, then per metric, then desktop versus mobile separately.

One projection worth making: the May 2026 dip and the steady mobile climb point the same way — the bar rises as the web gets faster, so a score that passes comfortably today can quietly slide toward the threshold as the field improves around it. Benchmarking is not a one-time pass/fail; it is a position you hold against a moving median. The practical discipline is to re-check field data quarterly, treat any new template or third-party tag as a potential regression, and never optimize the lab number you can see at the expense of the field number Google actually ranks on. The same field-data-first logic underpins broader technical-SEO work like crawl budget and log-file analysis, and CWV is one line item inside a full core update SEO audit.