Google tests Lighthouse for Agentic Browsing

google introduced an experimental 'agentic browsing' category in lighthouse to test how well websites support ai agents, focusing on machine-readability for future automated tasks. the audit checks support for google's webmcp api for logic exposure, the accessibility tree, visual stability, and the presence of llms.txt files. unlike standard audits, it outputs a ratio of passed checks rather than a numerical score. google explicitly stated that llms.txt is pointless for ai search, countering current optimization hype, and recommends semantic html to ensure agents can effectively navigate pages.