Lighthouse
Lighthouse is “an open-source, automated tool to help you improve the quality of web pages” (Google) — the measurement instrument beneath much of this wiki’s performance discussion, which had the metrics (core-web-vitals) but not the tool that audits them. Source: Chrome for Developers.
How it works
Point it at a URL; it “runs a series of audits against the page, then generates a report” with 0–100 scores. Runs from Chrome DevTools, a Node CLI, the Node module (programmatically), the web UI (PageSpeed Insights), or a browser extension. Performance metrics reported: FCP, Speed Index, LCP, TBT, CLS. The docs also keep a retired metrics list — Time to Interactive, First Meaningful Paint, First CPU Idle — which is where a metric this wiki cites from an older article should be checked against.
Five categories, and the fifth is not about humans
Re-read 2026-08-12. The categories are Performance, Accessibility, Best Practices, SEO and
Agentic browsing — “a set of deterministic audits” evaluating “how well your site is
constructed for machine interaction.” It checks WebMCP tool registration (declarative and
imperative), WebMCP schema validity, forms missing declarative WebMCP, the quality of the
accessibility tree for machine navigation, interactive-element naming, role and relationship validity,
[[llms-txt|llms.txt]] discoverability, and layout stability via CLS. It does not score 0–100 like
the other four — it reports a fraction, which the docs attribute to agentic web standards still being
in flux.
This page had four categories and was wrong when it was written, not merely out of date: the
category’s own documentation is dated 2026-05-05, a month before this page’s 2026-06-09 ingest.
The volatile flag caught an ingest error rather than drift in the world, which is the more useful
thing for it to have caught.
Two consequences for this wiki. First, CLS is now load-bearing in two different arguments — visual stability for a human reader (core-web-vitals) and visual stability so a machine can act on the right element. The same measurement, two audiences, and no source here yet says whether the thresholds should match. Second, the tool this spoke treats as its performance instrument has quietly become an AI-readiness instrument too, which puts lighthouse on a seam with llms-txt and generative-engine-optimization over in search-marketing-wiki.
WebMCP is now paged at webmcp in ../agentic-tooling-wiki, ingested the same day from the W3C
Web Machine Learning CG explainer and Chrome’s own docs. One fact from it belongs here: the audit
shipped ahead of the standard. WebMCP is a proposal in origin trial from Chrome 149, with no version
number and 108 open issues, and Lighthouse already scores pages against it. Every other category on
this page audits something settled.
Lab vs. field — the key caveat
Lighthouse is lab data: a controlled, single synthetic run — not real-user field data (the CrUX dataset behind core-web-vitals-wordpress-astro‘s rankings and Google’s actual ranking signal). The two routinely disagree, which matters for this wiki’s recurring tension: a great lab score and a poor field score can coexist. Use Lighthouse to diagnose (budgets can be enforced via Lighthouse-CI), CrUX to judge.
Related
core-web-vitals · performance-budget · page-weight · core-web-vitals-wordpress-astro · web-almanac · crux · webmcp · llms-txt · generative-engine-optimization