Google’s Lighthouse Tool Now Grades Your Agent-Ready Website

Google just gave every website a new report card, and this one isn’t written for humans. Chrome DevTools added a category to its Lighthouse testing tool called Agentic Browsing, and it checks whether AI agents can actually use your site, not just read it. If you run a business website, this changes what “optimized” means. An agent-ready website now needs to pass tests that have nothing to do with page speed or keyword density.

What the Agentic Browsing Audit Actually Checks

Chrome’s engineering team explained the update in a Chrome for Developers blog post on the agent-ready toolkit. The new category sits inside Lighthouse, the same free tool built into Chrome DevTools that developers already use to check performance and SEO. It’s available from Chrome version 150 onward, and you don’t need to flip any hidden settings. Open DevTools, run a Lighthouse report, and the category shows up automatically.

It tests three things. First, your accessibility tree, the structured map of your page that screen readers and, increasingly, AI agents use to understand what’s clickable and what isn’t. Second, your Cumulative Layout Shift score, because an agent that clicks a button right as it jumps somewhere else fails the task, the same way a person would. Third, whether your site supports WebMCP, a proposed standard that lets a website expose its forms and actions directly to an agent instead of forcing it to guess from the page layout.

Google made a point of not scoring this category the usual way. There’s no weighted average or letter grade. You get a pass or fail on each check plus a ratio, because the standards for agent-ready websites are still being written. Google is signaling direction, not handing out final grades.

Why AI Agents Are Already Browsing Your Site

This isn’t a hypothetical future problem. Chatbots and AI assistants already visit sites, and search itself has changed around them. We covered how Google AI Overviews became the default result for most searches, which means an AI summary often stands between your business and the click. Add agents that go a step further and actually complete a task on your page, booking a table, checking a price, filling out a contact form, and you’ve got a second audience that reads your site completely differently than a person does.

We’ve also written about why a well-structured website is what gets your business found by AI chatbots in the first place. The Agentic Browsing audit is the next logical step after that. Being found is one problem. Being usable once an agent lands on your page is another, and it’s the one most site owners haven’t thought about yet.

Not every business wants every bot crawling and acting on its behalf, either. If you’d rather control which AI systems get access, we broke down how Cloudflare now gives website owners granular AI bot controls. Being agent-ready and being selective about which agents you let in aren’t opposites. You can do both.

How to Make Your Website Agent-Ready

Start with the accessibility basics, because they carry double duty now. Give every button, link, and form field a clear programmatic name. If a “Book Now” button is just an icon with no label in the code, both a screen reader user and an AI agent will struggle with it. This is also core to how your website’s build decides your local SEO ranking, so fixing it helps you twice.

Next, look at layout stability. Popups, cookie banners, and ads that shift content around are already a UX problem. Now they’re also the reason an agent clicks the wrong thing and abandons the task. Run a Lighthouse report, check your CLS score, and fix whatever’s jumping around on load.

Finally, keep an eye on WebMCP. It’s still an emerging standard, and you don’t need to rush out and implement it this month. But if you’re planning a rebuild or a new feature that involves bookings, checkout, or account actions, ask your developer whether it makes sense to expose that flow through WebMCP so agents can use it directly instead of scraping your page.

You can run this audit yourself right now. Open Chrome DevTools on your own site, click the Lighthouse tab, and run a report. It only takes a minute, and it tells you exactly where you stand before a competitor gets there first.

Frequently Asked Questions

What is Lighthouse’s Agentic Browsing category?

It’s a new audit category in Google Chrome’s Lighthouse tool, available from Chrome 150 onward, that checks whether AI agents can reliably understand and interact with your website. It scores accessibility, layout stability, and support for the WebMCP standard.

Do I need to fix everything the audit flags right away?

No. Google built this category as informational rather than scored, since the standards are still developing. Treat it as a priority list rather than an emergency. Accessibility and layout stability fixes help you today regardless of AI agents, so start there.

What is WebMCP and do I need it?

WebMCP is a proposed standard that lets your website expose specific actions, like booking a slot or adding an item to a cart, directly to AI agents in a structured way. Most small business sites don’t need it yet, but it’s worth discussing with your developer if you’re planning a rebuild.

Getting Your Agent-Ready Website in Shape Before It’s a Standard

Google rolled this out quietly, as an informational category rather than a graded one, and that’s exactly why it’s worth acting on now. The businesses that clean up their accessibility tree, fix their layout shift, and start thinking about WebMCP today will have an agent-ready website long before this becomes a ranking factor. The ones that wait will be fixing the same issues later, under more pressure and with less runway.