How to Run a Lighthouse Agentic Browsing Audit (Step by Step)
You can score any URL against Google's agentic browsing category in about a minute with the Lighthouse CLI. Here is the exact command, how to read the result, and the gotchas to watch for.
Running the Audit in One Command
The agentic browsing category ships in Lighthouse 13.3.0 and runs from the command line with a single flag: --only-categories=agentic-browsing. You need Node.js and a Chrome install; no account or API key. For the full explanation of what it scores, see the Lighthouse Agentic Browsing guide.
That produces an HTML report you can share and a JSON file you can parse in automation.
How to Read the Result
The category header shows a fraction, not a 0–100 score — for example 1/3 or 3/3 — representing how many applicable checks pass. Below it, each audit shows pass, fail, warning, or Not Applicable.
In the JSON, the category score is a float from 0 to 1 (e.g. 0.34). Do not translate that into “34 out of 100” in a report — the shareable page headlines the fraction, so describe the result as “passes 1 of 3 checks” to stay consistent with what stakeholders see.
Three Gotchas to Watch For
1. Results fluctuate. Tool-registration timing, accessibility-tree construction, and late layout shifts can move the result between runs. Compare trends, not single runs.
2. Not Applicable is not failure. WebMCP audits show as Not Applicable until you adopt WebMCP; they are excluded from the ratio, so they do not lower your score.
3. It is experimental. The category and its rubric are still under development, so re-check after Lighthouse updates and qualify findings as “as of” a date.
Indexable's Technical SEO Agent Fixes This Automatically
The Technical SEO Agent audits every template against Google's agentic browsing checks and hands engineering a line-level fix sheet — CLS, accessibility tree, llms.txt, and WebMCP.
Frequently Asked Questions
Do I need an account or API key to run the audit?
No. The Lighthouse CLI runs locally with Node.js and Chrome. Use npx lighthouse@latest with --only-categories=agentic-browsing.
Why did my score change between two runs?
The agentic audits depend on dynamic conditions like tool-registration timing and layout shifts from late-loading content. Compare trends across several runs rather than trusting a single result.
Most Sites Pass 1 of 3 Agentic Checks. Yours Could Pass All Three.
Indexable's ten-agent system audits agentic browsing readiness across every template, fixes the gaps, and monitors your score as the standard evolves.
More in the Agentic Browsing Cluster
- Lighthouse Agentic Browsing — the full guide to Google's AI-agent readiness score (start here).
- The Three WebMCP Lighthouse Audits — registered-tools, form-coverage, and schema-validity.
- CLS and AI Agents — why Cumulative Layout Shift now decides whether an agent can click.
- The Accessibility Tree and AI Agents — the primary data model agents read your site with.
- llms.txt Explained — the machine-readable summary file agents read first.
- SEO vs AEO vs Agent-Readiness — the three stacked layers of AI visibility.
Where this fits your AI-visibility program: baseline your own site with the agent-readiness sprint, compare the platforms that automate these fixes in our ranked review of the best AI SEO agents, or get the executive case in AI SEO for CMOs.
Primary Sources Cited
- Google / Chrome Developers. Lighthouse Agentic Browsing scoring documentation. Published May 5, 2026. developer.chrome.com/docs/lighthouse/agentic-browsing/scoring