Your PageSpeed Insights score: what it counts and why it moves
Five lab metrics, weighted and blended into one grade, and why it is not the same as passing Core Web Vitals.
By ClaudeGPT Si Newsroom · Editorial team
· 3 min read
Your PageSpeed Insights score is a 0 to 100 grade from one Lighthouse lab test, worked out as a weighted average of five measured metrics. It is not the same thing as passing Core Web Vitals, and the easiest way to keep the two straight is seogeoaeo.ai's Core Web Vitals Snapshot, which reports them in separate sections for any public URL.
Where the number comes from
When you press Analyze, Google loads your page once in one of its data centers. On the mobile tab it pretends to be a mid-tier phone on a mobile network; on desktop it uses a wired connection. Lighthouse turns each metric into its own 0 to 100 score on a curve drawn from real-site data in the HTTP Archive, then blends them.
The five metrics and their weights
These are the weights in Chrome's Lighthouse scoring docs, as of November 2026:
| Metric | Weight | What it captures |
|---|---|---|
| Total Blocking Time | 30% | How long the page was too busy to respond |
| Largest Contentful Paint | 25% | When the biggest piece of content appeared |
| Cumulative Layout Shift | 25% | How much the layout moved |
| First Contentful Paint | 10% | When the first text or image showed |
| Speed Index | 10% | How fast the visible area filled in |
Three metrics carry 80% of the grade. If your score is low, heavy JavaScript (which drives blocking time) and a slow main image are the usual places to look first.
What the colors mean
- 90 to 100, green: good.
- 50 to 89, orange: needs improvement.
- 0 to 49, red: poor.
Do not chase 100. Google's docs call a perfect score extremely hard and not expected, and note that going from 99 to 100 takes about as much work as going from 90 to 94.
Why the score jumps around
Run the same page twice and the grade can change with no edits at all. Google lists the usual reasons: ads or A/B tests that change what loads, internet routing, the device and hardware doing the test, browser extensions, antivirus software and other work competing for the machine. Run a few tests and compare them instead of trusting one.
Mobile and desktop scores are also not comparable. Since Lighthouse 6, desktop has its own scoring curves.
Score versus the Core Web Vitals pass
The score is lab data: one load, one simulated device. The Core Web Vitals assessment above it is field data: real Chrome users over 28 days, judged at the 75th percentile. They can disagree, and Google's FAQ notes that good lab data does not necessarily mean real visitors are having a good experience. Lighthouse also cannot measure INP, since nobody clicks during a lab test, so blocking time stands in for it.
If you have to choose, the real-user pass is what visitors felt.
Raising it
The suggestions under the score do not count toward it directly; they help only when they improve a metric. Since Lighthouse 13 in October 2025, those suggestions appear as "insights", such as image delivery and render-blocking requests, and Google says the scoring did not change.
See the score and the pass together
The free PageSpeed Insights page shows the same score. What the seogeoaeo.ai snapshot adds is a cleaner read for repeat checks: real-user LCP, INP and CLS in one labelled section, the lab metrics and performance score in another, insights ranked by estimated savings, and tips in priority order you can verify after a fix. It runs in the browser, over MCP or through REST, so an agent like Claude Code can check the score after every change. Each run costs 20 credits.