Download websites that use web-vitals

About web-vitals

A performance library is one of the few things on a page that exists purely to watch the page. Nothing about web-vitals changes what a visitor sees. The library measures loading, responsiveness and layout stability in the real browser of a real visitor and reports those readings back to an endpoint the site owner controls. Detecting it is therefore not a finding about how a site was built, it is a finding about whether anyone is paying attention to how the site performs. As of October 2026 we detect web-vitals on 235 sites.

Google published the library in 2020 under the Apache 2.0 license, written in JavaScript, and the metrics it captures feed both search ranking signals and internal quality targets. A team that installs it has to do more than drop in a script: the readings need a destination, someone has to look at the dashboard, and somebody usually has to argue for the engineering time that follows a bad reading. The presence of a JavaScript libraries library of this kind is a proxy for a maturity level, which is why it correlates so strongly with the rest of the stack around it.

Version data shows how the library actually arrives. We record 23 distinct versions across the sites we can identify, and the release we see most often is 3. A spread that wide is not a story about deliberate upgrade discipline, because the library is usually pulled in by a framework, a commerce theme, or a tag manager template rather than installed by hand. The version on a site is closer to a fingerprint of when its build pipeline was last refreshed than to a choice anyone made, which makes the field useful for estimating how stale a front end is.

The honest caveat is bundling. When the library is compiled into an application chunk and minified, its traces can vanish entirely, so the sites we can confirm skew toward those loading it as a separate file or injecting it through a tag manager. The real installed base is larger than the count, and the gap is not random: sophisticated build setups are exactly the ones most likely to hide it.

Checking one domain answers a support question. The exported list answers a market question. Analytics and real-user-monitoring vendors use a population like this to size an audience that has already accepted the idea of field measurement, agencies selling performance work use it to find sites that are measuring but not fixing, and researchers use it to correlate instrumentation with frameworks, hosting, content delivery networks and commerce platforms across the whole crawl. None of those uses survive being done one lookup at a time, which is what the downloadable file is for.