Google PageSpeed Insights: how to read the report and what actually matters for search
Real-user data or the lab score, mobile or desktop, why the number changes on every run and which part Google uses for ranking. A practical guide to pagespeed.web.dev.
Performance · Published: 3 November 2026 · 6 min read · Анна Ковалёва
PageSpeed Insights is Google's free tool for measuring how fast a page is. Paste an address, wait about twenty seconds, and you get a big coloured number, a row of metrics and a long list of suggestions. Most people look at the big number first. It is the least important thing on the page.
Two reports in one
The report has two halves that come from completely different sources.
"Discover what your real users are experiencing" — the field data. It comes from the Chrome UX Report (CrUX): timings collected from real Chrome users who opted in to share usage statistics, over the last 28 days. Each metric is shown at the 75th percentile, the value three quarters of visits beat. On top sits the Core Web Vitals assessment: it passes when all three Core Web Vitals are in the "good" band.
"Diagnose performance issues" — the lab data. A single Lighthouse run from Google's servers, on an emulated mid-range phone with a throttled mobile connection (or a desktop profile on the other tab). This is where the 0–100 performance score lives.
When the field data is there, it is the truth about your visitors. The lab is a controlled experiment that helps you find out why.
The three numbers Google cares about
The Core Web Vitals and their "good" thresholds:
- LCP, Largest Contentful Paint — when the main content has loaded. Good: 2.5 s or less; poor: over 4 s.
- INP, Interaction to Next Paint — how quickly the page reacts to taps, clicks and key presses. Good: 200 ms or less; poor: over 500 ms.
- CLS, Cumulative Layout Shift — how much the content jumps. Good: 0.1 or less; poor: over 0.25.
The field section also shows FCP and TTFB. They are useful for diagnosis but are not part of the assessment. Each metric is covered in detail in Core Web Vitals.
Does it affect rankings?
Google uses Core Web Vitals as part of its page experience signals, and the data it uses is the field data — the same CrUX numbers shown at the top of the report. The lab score is not a ranking factor in itself. Google is also explicit that relevance comes first: a fast page does not outrank a more useful one, but between pages of similar quality, experience can tip the balance. The practical goal is a passed Core Web Vitals assessment on mobile, not a perfect 100.
Why there is no field data
"There is insufficient real-user data for this URL" is common for small or new sites. Chrome only reports a page once it has enough eligible visits. PageSpeed Insights then tries the whole origin (switch "This URL" to "Origin"), and if that is also too small, the lab run is all there is. That is not a penalty — it just means your numbers are estimates until traffic grows.
Reading the lab score
The performance score is a weighted blend of five lab metrics. In current versions of Lighthouse, Total Blocking Time carries 30%, Largest Contentful Paint and Cumulative Layout Shift 25% each, and First Contentful Paint and Speed Index 10% each. The bands are red below 50, orange from 50 to 89 and green from 90.
Three things to keep in mind:
- The score moves between runs. Network, server load and third-party scripts change from one run to the next; a swing of five points means nothing. Compare several runs, or the trend over weeks.
- Mobile is the one that matters. Google indexes the mobile version of a site first, and the mobile profile is far stricter. A page that scores 95 on desktop and 45 on mobile has a mobile problem.
- TBT is the lab stand-in for INP. A lab run does not click anything, so long JavaScript tasks during load are the closest proxy for slow interactions.
From suggestions to a plan
The "Opportunities" and "Diagnostics" lists are sorted by estimated saving, and each is tagged with the metric it helps. Work through them by metric:
- LCP: a faster server response, the hero image in AVIF or WebP at its displayed size with
fetchpriority="high"and neverloading="lazy", less render-blocking CSS and JavaScript. See image optimisation. - INP and TBT: fewer and lighter third-party scripts, long tasks split up, work deferred until after the first paint.
- CLS:
widthandheighton every image, a reserved slot for ads and embeds, fonts that do not reflow the text (web fonts). - Repeat visits: long cache lifetimes for static files — HTTP caching.
After deploying, remember the timing: the lab run reflects the change immediately, while the field data is a 28-day rolling window and catches up gradually.
PageSpeed in a SiteScan report
SiteScan calls the PageSpeed Insights API for your home page, with both the mobile and the desktop profile, and shows the result in the "PageSpeed Insights · home page" group next to the rest of the audit. Real-user data, when Google has it, appears under "Real users, last 28 days (CrUX)", and falls back to the whole site when the home page has too little traffic. So the speed numbers sit on the same page as the broken links, headers and SEO findings that often explain them — the lab and field difference is explained under data source. Run a free scan to see both.
Read next
Check how much of this is on your own site
A free scan surfaces broken links, JS errors and performance problems in about a minute.