Lighthouse Audit
An open-source, automated tool by Google for improving the quality of web pages.
Detailed Explanation
Lighthouse provides scores in five categories: Performance, Accessibility, Best Practices, SEO, and PWA. It identifies specific issues (like unoptimized images or missing meta tags) and provides actionable advice on how to fix them. It is built into Chrome DevTools and is the industry-standard tool for performance benchmarking.
Quick Summary
Lighthouse is Google's automated page audit tool, it grades Performance, Accessibility, Best Practices, SEO, and PWA, and explains specifically what to fix. It's lab data, useful for debugging, but the real scorecard is field data from real users.
Key Takeaways
- Five categories, each scored 0–100 with weighted sub-audits.
- Lab data: runs on a simulated 4G mobile device with throttled CPU, synthetic, reproducible, not real users.
- Field data (CrUX) is what Google actually uses for ranking; Lighthouse is for debugging.
- Available in Chrome DevTools, as a CLI (`npx lighthouse`), in PageSpeed Insights, and as a CI integration.
- Run multiple times, scores fluctuate ±5 points between runs due to throttling variance.
When to use it
- Pre-launch performance audits and reviewing PR-level regressions in CI.
- Diagnosing why LCP or INP is slow, Lighthouse points at specific resources or scripts.
- Quick accessibility wins (alt text, contrast, landmarks) catchable by axe rules.
- Tracking trends over time via Lighthouse CI to prevent silent regressions.
Common Mistakes
- Chasing 100/100 scores, diminishing returns past ~90 and the score doesn't directly impact ranking.
- Treating Lighthouse as the source of truth for Core Web Vitals; it's lab, not field.
- Running once and reacting; scores vary, so trend over multiple runs.
- Optimizing for Lighthouse while breaking real-user experience (lazy-loading the LCP image, for example).
Lighthouse Audit, Frequently Asked
Lighthouse vs. PageSpeed Insights?
PageSpeed Insights bundles a Lighthouse run *and* CrUX real-user data for the URL. Lighthouse alone is lab-only. For ranking-impact decisions, look at the CrUX section in PSI; for debugging specific bottlenecks, look at the Lighthouse section.
Should I run Lighthouse in CI?
Yes, with Lighthouse CI to track scores over time and alert on regressions. Don't fail builds on absolute thresholds (they're too flaky); fail on significant deltas vs. the baseline.