Audit any page — only what is actually wrong
Paste a URL. We look at the page once, run up to 53 checks, and report only what it actually gets wrong — each with why it matters and the fix.
Only what the page itself shows. Real Core Web Vitals are marked not measured, never guessed; a check that cannot be decided is skipped, not filled in.
Site files
Files the site publishes for machines: robots.txt, sitemap.xml and llms.txt.
Real performance (optional)
Runs a full Lighthouse pass in a real browser. Takes about 15 seconds and returns a large payload, so it is never automatic.
Open in PageSpeedAdvanced: check a saved copy
Some pages refuse automated visitors. If a page will not load here, save a copy yourself (View Source → select all → copy) and paste it below. The review runs inside this tab — nothing is uploaded.
What this tool does not do
- No backlink or rank data. Those need a crawl index we do not have, so we do not pretend.
- No guessed Core Web Vitals. LCP, CLS and INP need a browser; we label them not measured until you run Lighthouse.
- No guessing when a page hides itself. If a page builds itself in the browser and we cannot get the finished version, the affected findings say they are unproven instead of inventing a problem.
- No padding. If 200 checks fail, we say 200. We never invent findings to make a report look thorough.
How the checks are decided
Each check is a condition the page either meets or does not: something that must exist, a length limit, an ordering rule or a count. Thresholds come from Google's own documentation and are listed next to every finding. When a condition cannot be decided from what the page returns (for example layout shift), it is skipped rather than guessed.
Everything you see is produced in this tab. No account, no stored reports. If a page will not open directly, we may ask a reading service for a rendered copy — your URL is never stored. Privacy