TL;DR
Agentic browsing is a Lighthouse category, added in version 13.3 in May 2026, that scores how ready a page is to be read and acted on by AI agents rather than human visitors. It runs four checks: accessibility tree well-formedness, WebMCP manifest validity, llms.txt validity, and layout stability during agent interactions. Scored as a pass ratio (for example, 3 of 4), not a 0 to 100 score.
What is Agentic Browsing?
Agentic browsing is a Lighthouse category, introduced in Lighthouse 13.3 (May 7, 2026) and surfaced in PageSpeed Insights around May 21, 2026, that measures whether a web page is machine-readable and stable enough for AI agents to browse reliably.
Rather than a 0 to 100 performance-style score, the category is reported as a pass ratio (for example, 3 of 4). Google labels the category 'under development,' which means the exact checks inside it can change from one Lighthouse release to the next. See our Lighthouse Agentic Browsing news post for a fuller walkthrough.