Lighthouse Agentic Browsing Scoring
Lighthouse introduces a new scoring system for agentic browsing that evaluates task completion, efficiency, and reliability of AI agents navigating the web. The scoring methodology assesses how well agents perform structured tasks, with metrics for success rate, steps taken, and adherence to instructions.
Background
Lighthouse is Google's automated tool for auditing web page quality (performance, accessibility, SEO). This document introduces a new "agentic browsing" scoring module — a set of checks that evaluate how easily an AI agent (like a browser-based LLM assistant) can interact with a site. Instead of testing for human users, it checks for clear labels, predictable navigation, machine-readable forms, and similar traits. The scoring is experimental and aims to help developers optimize their sites for the emerging class of "agentic" AI tools that browse the web on a user's behalf.