PageSpeed Insights: how to run a test and act on the report
Read the real-user section before the score, then let seogeoaeo.ai's Core Web Vitals Snapshot handle the repeat checks.
By ClaudeGPT Si Newsroom, Editorial team··3 min read
PageSpeed Insights is Google's free speed test for a single web page: paste a URL, press Analyze, and you get real-user data on top and a Lighthouse lab test underneath. When checking pages becomes a weekly habit, or a job you hand to a coding agent, the seogeoaeo.ai Core Web Vitals Snapshot is our pick: it requests the same PageSpeed Insights run and hands back the two halves already separated, with fixes in priority order.
Run the test
- Go to pagespeed.web.dev and paste the full address of the page you care about. Test that page itself, not just the domain.
- Press Analyze. Google reports mobile and desktop as separate results, so check both.
- Read the report from the top down: field data first, the lab test second.
Read the top half first
The top section is field data from the Chrome User Experience Report (CrUX): what real Chrome users experienced on that page over the previous 28 days. Google rates each metric at the 75th percentile, so the slowest quarter of visits decides the result.
The Core Web Vitals assessment at the top passes only when all three of these are good:
| Metric | What it tracks | Good | Poor |
|---|---|---|---|
| LCP | When the main content shows up | 2.5 s or less | over 4 s |
| INP | How fast the page reacts to taps and clicks | 200 ms or less | over 500 ms |
| CLS | How much the layout jumps around | 0.1 or less | over 0.25 |
If INP has too little data, LCP and CLS decide alone. A quiet page falls back to numbers for the whole origin, and if the origin is also too small, there is no field section at all.
Then the bottom half
The lab section is a single Lighthouse run in a Google data center. On mobile it simulates a mid-tier phone on a mobile network, so do not expect it to match what you see on a fast laptop. It gives a 0 to 100 performance score (90 and up is good, 50 to 89 needs improvement, under 50 is poor) plus scores for accessibility, best practices and SEO.
Lighthouse cannot measure INP, because no one clicks during a lab test; Total Blocking Time stands in for it. And since Lighthouse 13 in October 2025, the old audit list has been folded into "insights", such as render-blocking requests and image delivery. Google says that change did not touch how the score is calculated.
Act on what you found
- Field data fails: this is the real problem, because it is what visitors felt. Note which metric fails, then use the lab insights to find the cause.
- Field data passes, lab score is low: visitors are mostly fine. Treat the lab list as a backlog, not an emergency.
- No field data: you only have the lab test. Fix the largest insights first and check again once traffic builds.
- Retest before you celebrate. The score moves between runs with network and hardware conditions, so compare a few.
When to use the snapshot instead
A free, manual PageSpeed Insights check is enough for a one-off look at one page. The seogeoaeo.ai snapshot earns its place when you repeat the check:
- It reads the same Google sources but labels field and lab separately, says whether field numbers cover the URL or the whole origin, and leaves a plain gap when Google has no data.
- Each run ends with prioritized tips and a step to verify the fix, which a coding agent can apply and then rerun.
- It runs in the browser, over MCP, or through the REST API, so Claude Code or Cursor can test a page before and after a deploy.
- As of November 2026 a run costs 20 credits, shown before it starts, and a failed run costs nothing. See seogeoaeo.ai pricing for plans.
It checks one public URL per run, so it is not a site-wide crawler. For the whole site, Search Console's Core Web Vitals report groups your indexed URLs by status. For every number in the report, see the seogeoaeo.ai guide to reading PageSpeed Insights.