Quick answer
- PageSpeed Insights shows two kinds of data: real-user data (Core Web Vitals from Chrome users) and a lab test (Lighthouse) with a 0 to 100 score.
- Google’s “good” thresholds: LCP 2.5 seconds or less, INP 200 milliseconds or less, CLS 0.1 or less.
- Passing Core Web Vitals matters more than hitting 100. A perfect lab score isn’t required to rank.
- The biggest wins: good hosting and caching, optimized images, fewer heavy scripts and a lightweight theme.
Chasing a perfect 100 in PageSpeed Insights can waste hours. Here’s what the score means and how to improve the parts that matter.
Understanding the report
| Section | What it shows |
|---|---|
| Core Web Vitals assessment | Real-user data over the last 28 days |
| Performance score | A lab test from Lighthouse, 0 to 100 |
| Opportunities and diagnostics | Specific improvements |
Core Web Vitals thresholds
| Metric | Measures | Good |
|---|---|---|
| LCP | Loading of the main content | 2.5 s or less |
| INP | Responsiveness to input | 200 ms or less |
| CLS | Visual stability | 0.1 or less |
How to improve your score
- Hosting and caching: fast servers and page caching (LiteSpeed Cache, WP Super Cache or W3 Total Cache, depending on your host).
- Images: resize, compress, use modern formats and lazy-load below the fold.
- Theme: use a lightweight theme. See fastest WordPress themes.
- Scripts: remove unused plugins and delay third-party scripts.
- CDN: serve files from near your visitors. See what is a CDN.
- Fonts: limit font files and preload key ones.
- Object caching: see Redis cache for WordPress.
Do you need 100/100?
No. Focus on passing Core Web Vitals for real users and on a fast, smooth experience.
FAQs
90+ is considered good in the lab test, but passing Core Web Vitals matters more.
Page experience is considered, but relevant content matters more.
The mobile test simulates a slower device and network.
LCP, INP and CLS: loading, responsiveness and visual stability.
Use your host’s recommended option; LiteSpeed Cache on LiteSpeed servers.
Lab tests vary slightly between runs.









