Agentic Browsing is the newest Lighthouse category. It checks whether AI agents, such as browser agents that read pages and fill in forms for a user, can understand and use a page. Google added it in Lighthouse 13.3 and labels it as still under development, so the checks may change.
Unlike the other categories, Lighthouse shows Agentic Browsing as a fraction of checks passed (for example 3/4) instead of a 0 to 100 score. Checks that do not apply to a page, such as WebMCP checks on a site without WebMCP, are left out.
What it checks
| Check | What passes |
|---|---|
| Accessibility tree | The page has a well-formed accessibility tree: semantic landmarks, headings, and named buttons, links and form fields. |
| llms.txt | If the site serves /llms.txt, it follows the recommended format: a title, a short summary and lists of links. |
| Agent resource discovery (ARD) | If the site publishes an ARD manifest describing resources for agents, it is valid. |
| WebMCP schemas | If the page registers WebMCP tools, their input schemas are valid. |
| WebMCP form coverage | Lists forms that could get WebMCP annotations (informational). |
| Layout stability | Cumulative Layout Shift is low, so elements do not move while an agent acts on them. |
How to improve it
Most fixes overlap with accessibility work: use real buttons and links, label every form field, and keep the layout stable. Serve a well-formed llms.txt if you want AI tools to find your key pages. Only add WebMCP or an ARD manifest if you use them, and validate them when you do.
AiroPages runs the Agentic Browsing checks on every scan and shows the result for each site, next to the four classic categories.