Google Lighthouse has added a new 'Agentic Browsing' category that audits how well websites work for user-facing AI agents. Unlike other Lighthouse categories, it provides pass/fail diagnostics rather than a 0-100 score. The audits cover four areas: presence of an llms.txt file (a robots.txt equivalent for AI agents), WebMCP integration (a Chrome Origin Trial API for explicitly exposing site actions to agents), accessibility tree quality (semantic HTML and ARIA labels that agents rely on for navigation), and Cumulative Layout Shift (CLS) stability to prevent agents from clicking wrong elements. The category is designed to be integrated into CI/CD pipelines as a quality gate on every deploy, treating agentic readiness as a first-class metric alongside performance and accessibility.
Table of contents
The Agentic Browsing auditllms.txtWebMCP integrationAccessibility for agentsLayout stabilityPutting it into a CI pipelineWhere this fits alongside DevTools for agents180.7K Impressions7 Comments