Check Your Agentic Readiness

Agentic browsing is when AI agents (like Chrome's built-in agent, ChatGPT operators, or autonomous assistants) navigate and act on websites on a user's behalf. Chrome Lighthouse added an Agentic Browsing category that measures how readable and actionable your page is to these agents, using the Chrome DevTools Protocol and the live accessibility tree.

WebMCP is a browser standard that lets a page expose structured tools to AI agents. Declarative tools annotate existing forms and navigation with attributes like data-toolname and data-tooldescription, while imperative tools register actions in JavaScript via navigator.modelContext or document.modelContext. This tells agents exactly what actions a page supports and what parameters they take.

Lighthouse runs in a live Chrome instance and reports a fractional score (the ratio of agentic-readiness checks passed) using the DevTools Protocol WebMCP domain and the rendered accessibility tree. This tool fetches the page source from the URL you enter, so it cannot evaluate runtime-registered tools, computed accessibility names, or layout shifts that only happen after rendering. Treat this checker as a fast source pre-check, not the authoritative score.

Start by exposing WebMCP tools on your highest-intent forms and navigation, give every tool a clear name and description, and ensure each input has an accessible label. Then fix accessibility tree basics (html lang, image alt text, one h1, semantic landmarks), set explicit width and height on media to avoid layout shift, and publish an llms.txt at your domain root. Finally, run the official Chrome Lighthouse Agentic Browsing audit to confirm.