Website speed: server responses and user experience
A fast HTML response does not guarantee that visitors quickly see the main content. Images, fonts, scripts, third-party components and layout movement all affect the experience. This tool separates HTML responses, Lighthouse simulations and real-user Core Web Vitals so operators and developers can work from the same evidence.
Look beyond red, amber and green: check the source
Each metric card shows a value, status and explanation, using icons and text alongside color. Gray means not measured. Lab ratings use Lighthouse scores returned by Google; field assessment uses the 75th percentile of LCP, INP and CLS. Missing data is never guessed.
Frequently asked questions
Which metrics does this website speed test check?
The professional test uses Google PageSpeed Insights for Lighthouse LCP, CLS, FCP, TBT, Speed Index and a performance score. When Google returns CrUX data, real-user LCP, INP and CLS are shown separately. The lightweight test measures an HTML request from our server.
Does a green lab score mean Core Web Vitals passed?
A green lab score describes only that simulated run. A Core Web Vitals pass requires good real-user LCP, INP and CLS at the 75th percentile. Missing field metrics remain unknown, and TBT never replaces INP.
Why is real-user data missing?
Public CrUX coverage depends on sample size, discoverability and upstream data availability. Missing data is explicitly marked as not measured. Origin-level data is labeled and is never presented as URL-level experience. Check official PageSpeed or Search Console for more context.
How do mobile and desktop tests differ?
Lighthouse uses device emulation, network conditions and device-specific scoring curves. Compare changes within the same device mode and keep the test time and version. A single fluctuation does not establish a lasting change.
Is HTML response time TTFB or full page load time?
No. The lightweight measurement includes DNS, connection, redirects and HTML download from our server. It does not execute scripts, load images or fonts, or render the page. It is not isolated TTFB, full browser load time or CWV.
How can I use ChatGPT, Claude or Doubao to improve the results?
At the top of the report, copy the evidence and optimization prompt, then open an AI provider and paste it yourself. Reports are not sent automatically. Review the content first, verify proposed changes, and retest with the same device mode.
What happens if Google fails or returns a quota error?
Failures, timeouts and quota errors are shown explicitly. Estimated data is never substituted for Lighthouse. You can run the separate HTML response test or open official PageSpeed; the different methods remain clearly labeled.