WebMCP Lighthouse Score: What Is the Agentic Browsing Category?

What changed in Lighthouse 13.3

Google shipped Lighthouse 13.3 with a new audit category: Agentic Browsing. It sits alongside Performance, Accessibility, Best Practices, and SEO in the Lighthouse report panel in DevTools, and Google's own docs note the category is "still under development and subject to change."

The category scores how ready a page is for AI agents, not just search crawlers. It's a fraction score built from six equally-weighted audits, not a traditional 0–100 points system.

What the Agentic Browsing audits check

The category runs six audits, each counted equally toward the score:

WebMCP tools registered

Lists the WebMCP tools registered on document.modelContext at the time of analysis.

WebMCP schemas are valid

Checks that each tool's inputSchema is well-formed. AI agents need a valid schema to understand and call a tool correctly.

WebMCP form coverage

Flags forms on the page that have no matching WebMCP annotation, so agents can identify and fill them reliably instead of guessing at field names.

Accessibility tree is well-formed

A well-formed accessibility tree helps AI agents navigate and interact with the page independently of your WebMCP tools.

Cumulative Layout Shift

The standard CLS metric. A page that shifts under an agent's feet is as much of a problem for agents as it is for human visitors.

llms.txt follows recommendations

Checks that your llms.txt file exists, is a Markdown file with at least one H1 header, and contains links.

Why this score matters now

Google hasn't said whether Agentic Browsing scores will ever factor into search ranking, and the category is explicitly marked as experimental. But Core Web Vitals followed the same arc: announced as informational, then folded into ranking years later. Betting on the same pattern isn't guaranteed, but it's a reasonable one.

Right now, the score is purely informational. Getting ahead of it costs little, since the underlying work (registering tools correctly, keeping a valid llms.txt, avoiding layout shift) is worth doing for AI agents today regardless of whether it ever touches ranking.

How to check your current score

Open Chrome DevTools, run Lighthouse, and select the Agentic Browsing category if it's available in your Chrome version. You'll see all six audits with pass, fail, or not-applicable status.

The Lighthouse CLI works too, and doesn't require a specific Chrome version for the non-WebMCP audits:

npx lighthouse https://yoursite.com --only-categories=agentic-browsing --output=html

The three WebMCP-specific audits rely on a DevTools Protocol domain that only exists in Chrome builds that ship the native WebMCP implementation. On an older Chrome they'll show as not applicable rather than pass or fail.

How Aigentably helps your score

When you paste the Aigentably script tag, it:

Fetches your tool definitions from the Aigentably API
Registers them on document.modelContext with valid JSON Schema input schemas
Loads the WebMCP polyfill for browsers without native support

That covers the tool-registration and schema-validity audits without manual work. Form coverage, accessibility, CLS, and llms.txt depend on how your page is built, Aigentably's dashboard now runs a real Lighthouse Agentic Browsing scan against your site weekly and shows the score with a breakdown per audit, so you can see exactly where to improve instead of guessing.