AI Agent Readiness Check

What is agentic browsing?

An AI agent is software that visits a website to accomplish something on a person's behalf — finding an opening time, comparing prices, filling in a booking form. It reads the same page a person does, but it cannot glance at a layout and infer what a thing is. It depends on the structure underneath: headings, labels, landmarks, and the accessibility tree.

Google added an Agentic Browsing category to Lighthouse to measure that. This tool runs that audit on your page and shows the result. Google's own documentation for the checks is here.

What the checks cover

Google may add to this category over time, so your results may show checks not listed here. Broadly it covers two things:

• Can an agent read the page? Whether the accessibility tree is well-formed, and whether the layout is stable. Content that moves while loading (measured as Cumulative Layout Shift) breaks a human's reading and an agent's task in the same way — it acts on something that is no longer where it was.
• Can an agent use the page? WebMCP is an emerging way for a site to expose its actions — search, book, add to basket — so an agent can call them directly instead of guessing at form fields. These checks validate such an integration if you have one.

Why most sites see "not applicable"

Most of the WebMCP checks will not apply to your site, because very few sites have adopted it yet. That is not a failure and there is nothing to fix. The score you see counts only the checks that actually apply to your page.

If you want to improve the checks that do apply, they are ordinary quality work rather than anything agent-specific: proper heading order, labelled form fields, semantic landmarks, and a layout that does not shift while it loads. The same work improves accessibility for people using screen readers, which is the more immediate reason to do it.

Is this the same as blocking AI crawlers?

No, and the distinction matters. robots.txt governs whether crawlers like GPTBot and ClaudeBot may fetch your pages for training and retrieval — you can check that with our robots.txt analyser. Agentic browsing is about whether an agent that is already on your page can understand and use it. A site can allow every crawler and still be unusable to an agent, or block them all and score perfectly here.