What PageSpeed Insights Actually Checks
PageSpeed Insights runs four categories of audits with dozens of checks each. Here's what every one of them measures.
Table of contents
- The four categories
- Performance (100 points)
- LCP — Largest Contentful Paint (25%)
- TBT — Total Blocking Time (30%)
- CLS — Cumulative Layout Shift (25%)
- FCP — First Contentful Paint (10%)
- Speed Index (10%)
- INP — Interaction to Next Paint (0% in score, measured separately)
- Plus "Opportunities" and "Diagnostics"
- Accessibility (100 points)
- Color and contrast
- Names and labels
- Navigation
- ARIA
- Best practices
- Best Practices (100 points)
- Security
- Browser console
- User experience
- Trust and safety
- SEO (100 points)
- Lab data vs Field data
- Field data (top of the report)
- Lab data (lower in the report)
- How to use the report
- The bottom line
PageSpeed Insights runs a full Lighthouse audit on every URL you submit. There are four categories with dozens of individual checks. This is what each one actually measures.
The four categories
| Category | Max score | What it measures |
|---|---|---|
| Performance | 100 | Loading speed, Core Web Vitals, blocking work |
| Accessibility | 100 | Automated checks against WCAG 2.1 |
| Best Practices | 100 | Security, modern APIs, console errors |
| SEO | 100 | Crawlability, basic on-page SEO |
The scores are independent — you can have a perfect SEO score and terrible Performance, or vice versa.
Performance (100 points)
The Performance category is the headline number most people obsess over. It's a weighted average of six metrics:
LCP — Largest Contentful Paint (25%)
Time until the largest visible element loads. Good ≤ 2.5s. See LCP explained.
TBT — Total Blocking Time (30%)
Time the main thread was blocked > 50ms during page load. Good ≤ 200ms. The biggest weight in the score.
CLS — Cumulative Layout Shift (25%)
Sum of all unexpected layout shifts. Good ≤ 0.1. See CLS explained.
FCP — First Contentful Paint (10%)
Time until any visible content appears. Good ≤ 1.8s.
Speed Index (10%)
How quickly the visible area is populated, frame by frame. Good ≤ 3.4s.
INP — Interaction to Next Paint (0% in score, measured separately)
Worst case input-to-paint time across the session. Good ≤ 200ms. Reported by Lighthouse but not part of the lab score because it requires real interactions. See INP explained.
Plus "Opportunities" and "Diagnostics"
Below the score, you get specific recommendations:
- Opportunities: Things you can change with estimated time savings (e.g. "Properly size images — save 1.2s")
- Diagnostics: Observations about your page (e.g. "Avoid an excessive DOM size")
- Passed audits: What you did right
Accessibility (100 points)
Lighthouse runs roughly 50 automated accessibility checks. Common ones:
Color and contrast
- Background and foreground colors have sufficient contrast ratio (4.5:1 for normal text, 3:1 for large text)
Names and labels
- Buttons have an accessible name
- Form elements have labels
- Image elements have
altattributes - Links have discernible text (not just "click here")
- Document has a
<title>
Navigation
- Heading elements appear in sequential order
- Page has a logical tab order
- Skip-link or focus indicators where needed
ARIA
- ARIA attributes are valid and used correctly
aria-*attributes match their roles
Best practices
[user-scalable="no"]is not used (lets users zoom)<html>element has alangattribute- Lists use
<ul>,<ol>, and<li>correctly
Critical caveat: Lighthouse only catches about 30% of accessibility issues — the ones machines can detect. Things like keyboard usability, screen reader experience, and meaningful copy for screen readers all require manual testing. A 100 Accessibility score is necessary but not sufficient. See website accessibility score explained.
Best Practices (100 points)
Best Practices covers a grab bag of modern web hygiene. Major checks:
Security
- Uses HTTPS
- Avoids deprecated APIs (e.g. AppCache)
- No mixed content (HTTP resources on HTTPS pages)
- Has a Content Security Policy
- Doesn't use known vulnerable JavaScript libraries
Browser console
- No browser console errors
- No browser deprecation warnings
User experience
- Allows users to paste into password fields
- Doesn't request notifications or geolocation permission on page load (only after user action)
- Document has a valid
<!DOCTYPE html>
Trust and safety
- Doesn't use unload event listeners (prevents back/forward cache)
- Defines
charsetearly
See best practices score explained for details.
SEO (100 points)
SEO checks technical SEO basics — see what is a good SEO score for the full list. Main categories:
- Page has a
<title>and meta description - Page returns 200 status code
- Page isn't blocked from indexing
- Has a valid
robots.txt - Mobile viewport is set
- Tap targets are appropriately sized
- Images have alt text
- Document has a
langattribute - Links are crawlable (use
<a href>) - Structured data is valid (if present)
Lab data vs Field data
PageSpeed Insights shows two distinct sections:
Field data (top of the report)
Real-user measurements from the Chrome User Experience Report (CrUX). This is data from actual Chrome users visiting your site over the last 28 days. This is what Google uses for ranking. You'll only see this if your site has enough traffic.
Lab data (lower in the report)
A single Lighthouse audit run on Google's servers right now, with standardized throttling. Useful for diagnosing specific issues. Variance is high — averaging 3–5 runs gives a more reliable number.
If field data is great but lab data is bad, your audit is probably running on a different version of your site than real users see (e.g. logged-in vs logged-out, A/B test variant). If lab is great but field is bad, real-world conditions (CDN, geography, device class) are causing problems.
How to use the report
- Look at field data first. If your CWV are green for real users, you're good for SEO regardless of the lab score.
- Use lab data to debug. When something fails, the Opportunities and Diagnostics sections tell you what to fix.
- Don't chase 100. A 90 with no quick wins left is better than a 95 you got by removing essential features.
- Measure after every change. Run a free audit and confirm your fixes worked.
The bottom line
PageSpeed Insights is four audits in one: Performance, Accessibility, Best Practices, and SEO. Each measures a different dimension of site health. The Performance score combines six timing metrics, with TBT, LCP, and CLS doing most of the weight. Field data drives rankings; lab data drives debugging.
For the broader Lighthouse picture, see how to read a Lighthouse report.
Frequently Asked Questions
How is the PageSpeed Insights performance score calculated?+
It's a weighted average of six metrics — LCP (25%), TBT (30%), CLS (25%), FCP (10%), Speed Index (10%). Each metric is scored on a log-normal curve against real-world data.
Why does my Accessibility score not equal 100 even though I have no errors?+
Lighthouse only flags issues it can detect automatically — about 30% of accessibility problems. A perfect Lighthouse Accessibility score doesn't mean your site is fully accessible. Manual testing is still required.
What's the difference between Opportunities and Diagnostics?+
Opportunities show specific things you can change with an estimated time savings. Diagnostics are observations about your page that may or may not need action. Opportunities are usually the higher-impact list.
See how your site really performs
Run a full website health check on mobile and desktop in 30 seconds — no signup needed.