Google Lighthouse Adds a New Agentic Browsing Category to Test AI‑Ready Websites

Google’s Lighthouse, the popular open‑source performance and accessibility audit tool, has just expanded its feature set. The new Agentic Browsing category evaluates how well a site is prepared for interaction with AI agents—software that can read, understand, and act on web content. This addition reflects Google’s growing focus on the “agentic web,” a future where intelligent assistants can navigate sites, fill out forms, and even complete purchases without human intervention.

What Is Agentic Browsing and Why Does It Matter?

Traditionally, Lighthouse has measured page speed, accessibility, best practices, and SEO. The Agentic Browsing category shifts the lens toward machine‑centric usability. It asks: Can an AI agent reliably locate the information it needs, interpret the page’s structure, and perform actions such as clicking a button or submitting a form? If a site scores poorly, it may appear in search results or be used by AI assistants, but the experience could be confusing or broken for the underlying technology.

Google labels this new category as experimental. That means it is only available in Chrome Canary, the bleeding‑edge channel of the browser that receives daily updates. Marketers and developers who want to run the audit must install Canary and use the Lighthouse panel within it.

How Lighthouse Audits AI Agent Readiness

The Agentic Browsing audit consists of three distinct checks, each targeting a different aspect of machine interaction:

• WebMCP (Web Machine‑Readable Content Protocol) – Verifies that your site implements the WebMCP standard, which provides structured instructions for AI agents on how to navigate and act.
• Agent‑Centric Accessibility – Examines the portion of the accessibility tree that AI agents rely on. It checks for proper naming, labeling, tree integrity, and visibility of interactive elements.
• Stability and Discoverability – Measures layout stability via Cumulative Layout Shift (CLS) and looks for an llms.txt file, a file that signals to AI agents what content is available for machine consumption.

Unlike the classic Lighthouse score that ranges from 0 to 100, the Agentic Browsing result is expressed as a ratio: the number of checks passed divided by the total number of checks. For example, a site that passes two out of three checks would display a 2/3 score.

Implications for Marketers and Web Developers

Google’s push toward an agentic web is not just a theoretical exercise; it has concrete business implications. Several new tools and features are already in place or on the horizon:

• AI Performance Report in Merchant Center – Tracks how AI‑generated shopping experiences perform, providing insights into conversion rates and user satisfaction.
• AI Assistant Channel in GA4 – Segments traffic that originates from AI assistants, allowing marketers to attribute conversions and engagement to these new channels.
• Google‑Agent User Agent – A dedicated crawler identity that signals when Google’s AI agents are visiting a site, enabling site owners to monitor and optimize for these visits.
• Universal Commerce Protocol – A standard that lets AI agents complete purchases across multiple retailers without leaving Google’s ecosystem.
• WebMCP (again) – A proposed standard that, if adopted, would let sites explicitly direct AI agents on how to act, from clicking a button to filling out a form.

For marketers, the Agentic Browsing audit is a warning sign: if your site is not ready for AI agents, you risk losing visibility in AI‑driven search results and commerce experiences. For developers, it is an opportunity to future‑proof sites by ensuring that machine‑readable content, accessibility, and layout stability are up to par.

Getting Started with Agentic Audits

Here’s a quick step‑by‑step guide to run the new audit:

1. Install Chrome Canary from the official Chromium website.
2. Open the site you want to test.
3. Open the Chrome DevTools (Ctrl+Shift+I or Cmd+Opt+I).
4. Navigate to the