Google Lighthouse: Run and Interpret a Page Audit

Google Lighthouse is an open-source automated tool for auditing one web page at a time. It reports on performance, accessibility, SEO and other quality checks. Use failed audits as diagnostic leads, reproduce them under controlled conditions and confirm user impact before prioritising work.

What Is Google Lighthouse?

Chrome's Lighthouse documentation explains that it can audit public or authenticated pages from Chrome DevTools, the command line or a Node module. Each report reflects one URL, one run and its test environment. It is not a site crawler.

How to Run an Audit

1. Open the exact page and state you want to test in Chrome.
2. Open DevTools, select Lighthouse and choose the relevant categories and device mode.
3. Reduce background activity and extensions, then run the audit more than once.
4. Save the report with the URL, time, Lighthouse version and test configuration.
5. Open failed audits and verify the affected element or request in DevTools.

The CLI command is lighthouse <url>. Pin versions in automated workflows so tool changes do not masquerade as site regressions.

Interpret the Performance Score

The performance score is a weighted lab result under the run's simulated conditions. It can change because of CPU, network, page variability or Lighthouse updates. Compare medians from repeated, similarly configured runs and inspect the metric traces rather than chasing a single number.

Core Web Vitals are LCP, INP and CLS. Google's Web Vitals guidance states that Lighthouse cannot measure INP in a lab run because there is no real user interaction; Total Blocking Time is a lab proxy. Use PageSpeed Insights, Search Console or your own real-user monitoring for field data.

Interpret the Accessibility Score

The accessibility category finds automatable conditions such as missing accessible names, some contrast failures and structural problems. A high score does not establish WCAG conformance. Test keyboard operation, focus order, zoom, screen-reader output, error handling and content meaning manually. WAVE can provide a second automated view.

Interpret the SEO Score

The SEO category checks a limited set of technical basics that Lighthouse can observe on the tested page. It does not assess search demand, content usefulness, internal-link architecture, backlinks or whether Google selected the same canonical. Use a site-wide audit checklist around it.

Interpret Best Practices

Review each failed check against the current browser documentation and reproduce it in DevTools. The category is a useful development prompt, but it is neither a penetration test nor a complete privacy or security review.

Use Lighthouse CI for Regressions

Lighthouse CI can run Lighthouse in a development workflow and compare results. Select representative templates, pin the environment as far as possible and set budgets from a reviewed baseline. Treat a failed threshold as a signal to investigate, not automatic proof of a user-facing regression.

Limitations

Lighthouse observes one page state in a lab environment and only automates checks it knows how to evaluate. It cannot replace field performance data, assistive-technology testing, a site crawl, business context or human review. Store its report as one evidence source within the wider audit.

Technical guidance checked against official Chrome and web.dev documentation on 11 September 2026.