Google has told site owners for years to get their Core Web Vitals into the green. How many of the web’s biggest sites actually do? We pulled Google’s own real-user measurements, the Chrome UX Report, for the top 1,000 websites. For teams that track search performance, the answer is a useful reality check.
Core Web Vitals are three measurements of what real visitors experience. Google considers a site good when, at the 75th percentile of visits, all three meet these thresholds:
| Metric | What it measures | Good |
|---|---|---|
| Largest Contentful Paint (LCP) | How long the main content takes to appear | 2.5 s or less |
| Interaction to Next Paint (INP) | How quickly the page responds to a tap or click | 200 ms or less |
| Cumulative Layout Shift (CLS) | How much the page jumps around while loading | 0.1 or less |
The headline numbers
Of the top 1,000 domains in the Tranco ranking, 555 had enough Chrome traffic for Google to publish data; most of the rest are infrastructure domains that people never visit directly.
| Phone | Desktop | |
|---|---|---|
| Pass all three | 34.3% | 60.0% |
| Good LCP | 56.1% | 76.2% |
| Good INP | 51.4% | 90.9% |
| Good CLS | 83.0% | 79.6% |
| Median LCP | 2.37 s | 1.85 s |
| Median INP | 196 ms | 82 ms |
| Median time to first byte | 908 ms | 678 ms |
Two in three top sites fail on phones, where most people now browse. On desktop, a majority pass.
Responsiveness is the problem
On phones, INP fails more often than anything else. Only half of top sites respond to taps within 200 milliseconds at the 75th percentile, against nine in ten on desktop. The median site sits at 196 ms, right on the edge.
| What failed (phone) | Sites |
|---|---|
| LCP and INP | 114 |
| INP only | 91 |
| LCP only | 57 |
| All three | 48 |
| LCP and CLS | 18 |
| CLS only | 17 |
INP replaced First Input Delay as a Core Web Vital in March 2024, and it is harder to pass: it counts every interaction in a visit, not just the first. Heavy JavaScript, third-party tags and slow phones all show up in it. Layout shift, by contrast, is mostly solved: 83% of top sites pass on phones.
Bigger is not faster
The most visited sites did slightly worse, not better.
| Rank | Sites with data | Pass on phone |
|---|---|---|
| Top 100 | 48 | 27.1% |
| 101 to 500 | 198 | 34.8% |
| 501 to 1,000 | 293 | 35.2% |
Big names, phone results
| Passed | Failed |
|---|---|
| Amazon, Wikipedia, LinkedIn, Reddit, eBay, PayPal, Spotify, Etsy, Temu, CNN, The New York Times, BBC, The Guardian, Bing, Yahoo | Google, YouTube, Facebook, Instagram, Microsoft, Apple, Netflix, X, TikTok, GitHub, ChatGPT, Booking.com, Shopify, Canva, Pinterest, Airbnb |
Some failures are narrow. Google’s search home page loads its content in 1.1 seconds but responds to taps at 228 ms, just over the line. Netflix responds to taps in time, at 199 ms, but its main content takes 2.7 seconds to appear, just over the 2.5-second line. Others are not close: YouTube’s main content takes 5.9 seconds to appear on phones at the 75th percentile, and ChatGPT takes over a second to respond to a tap.
What this means for SEO
Google says: “We highly recommend site owners achieve good Core Web Vitals for success with Search”, and that, with other page experience aspects, they align “with what our core ranking systems seek to reward”. The data shows how much room there is. If two-thirds of the biggest sites fail on phones, passing is a real way to stand out, especially against competitors in the same results.
Practical priorities from the numbers:
- Start with INP on mobile. It is the most common failure. Audit long JavaScript tasks, third-party scripts and what runs on the first tap.
- Then LCP. Slow servers are part of it: the median top site takes 908 ms just to deliver the first byte on phones.
- Measure field data, not just lab scores. These numbers come from real Chrome users. A fast test on an office connection can hide a slow experience on a mid-range phone abroad.
- Check by market. Visitors far from your servers see slower pages. Our guide to testing website speed from different countries shows how to measure what each market sees, using residential proxies to load pages as local visitors.
How we did it
On 19 November 2026 we queried Google’s Chrome UX Report API for the top 1,000 domains in the Tranco ranking (list Q2K34), the same list we used for our IPv6 study. For each domain we requested the www origin, then the bare domain, separately for phones and desktops. The data covers real Chrome users over the 28 days to 17 November 2026, at the 75th percentile. A site passed if LCP, INP and CLS were all good; sites without INP data were left out of the pass rate, leaving 539 on phones and 538 on desktop. The data is per origin, so it blends every page on a site.
The bottom line
Only a third of the web’s biggest sites pass Core Web Vitals on phones, and the top 100 do worse than the rest. Responsiveness is the main reason: half of top sites are too slow to react to a tap. Desktop is in much better shape. For anyone competing in search, a fast, responsive mobile site is still a way to beat much bigger names.
Sources and references
- Google Search Central, Understanding Core Web Vitals and Google search results: developers.google.com
- web.dev, Web Vitals: web.dev
- Chrome UX Report API: developer.chrome.com
- web.dev, Interaction to Next Paint becomes a Core Web Vital (March 2024): web.dev
- Tranco, list Q2K34: tranco-list.eu