How the Website Speed Test works

90 or above is Good, 50–89 Needs improvement, and below 50 is Poor. But the score is a lab summary — Google’s ranking systems look at the three Core Web Vitals from real-user data, so a page can rank fine with a mediocre score if its real-user LCP, INP and CLS pass.

The mobile test simulates a mid-range phone with a 4x CPU slowdown on a slow 4G connection, while desktop assumes fast hardware and broadband. Most real traffic is mobile, which is exactly why Google tests it under pressure — and ranks you on it.

Lab data is one simulated page load run by Lighthouse — controlled and repeatable, good for debugging. Field data is the 75th-percentile experience of real Chrome users over the last 28 days (the CrUX dataset), and it is what Google actually uses to assess Core Web Vitals.

Field data comes from the Chrome UX Report, which only publishes a page once enough real Chrome users have visited it over the trailing 28 days. Below that threshold, PageSpeed Insights falls back to origin-level data — real users across the whole site rather than this specific page — and if the origin is too small too, there is no field data at all, so the report falls back to this run’s lab metrics instead. Interaction to Next Paint can only ever come from field data, so a page with no field data also has no INP reading.

Yes — Core Web Vitals are a confirmed ranking signal, assessed from field data at the 75th percentile. The effect is modest compared with content relevance, but speed also compounds through lower bounce rates and higher conversion, which is usually where the real money is.

Run-to-run variance is normal: network conditions, server response times, third-party scripts, rotating ads and A/B tests all shift the numbers. A few points of movement means nothing; consistent movement across several runs is a real change.

Google’s three user-experience metrics: Largest Contentful Paint (loading — Good at ≤2.5s), Interaction to Next Paint (responsiveness — Good at ≤200ms) and Cumulative Layout Shift (visual stability — Good at ≤0.10). A page passes when all three are Good at the 75th percentile of real visits.

Yes — every embedded script adds its own network request and runs JavaScript on the same main thread as your own code, which is exactly what Total Blocking Time and Interaction to Next Paint measure. A single heavy ad network or chat widget can cost more than the rest of the page combined. Lighthouse’s audits already account for this: when a third-party script is expensive enough to be worth fixing, it can appear in the top-five opportunities like any other audit, sorted by estimated saving.