Lighthouse Now Scores How AI-Ready Your Website Is

Google's Lighthouse — the tool developers have used for years to audit site performance, accessibility, and SEO — just added a new scoring category: Agentic Browsing. It evaluates how well your website is built for machine interaction. Not human interaction. Machine.

This is a quiet but significant signal about where the web is heading.

What the Agentic Browsing Score Actually Measures

Unlike Lighthouse's 0–100 performance scores, the Agentic Browsing category doesn't produce a single weighted number. It returns a pass/fail ratio — how many agentic readiness checks your site passes out of the total — alongside specific audit results flagging errors or warnings where your site falls short.

The audits focus on three areas. First, WebMCP integration: whether your site uses the WebMCP API to explicitly expose its tools, forms, and logic to AI agents. Second, accessibility tree integrity: whether every interactive element has a proper programmatic name, valid roles, and correct parent-child relationships — because AI agents navigate your site the same way screen readers do, through the accessibility tree, not the visual layer. Third, stability: whether your layout shifts enough during load that an agent targeting an element finds it has moved before it can interact with it.

There's also a check for llms.txt — a machine-readable summary file at your domain root that tells AI agents what your site does and how to use it.

The Accessibility Tree Is Now a Business Asset

Here's the part most marketing and growth leaders haven't fully absorbed yet: AI agents don't see your website the way humans do. They don't register your hero image, your font choice, or your color palette. They parse the accessibility tree — a structured representation of your page's interactive elements, labels, and relationships.

If your site has unlabeled buttons, broken ARIA roles, or interactive elements hidden from the accessibility tree, a human visitor might navigate around it without noticing. An AI agent hits a wall.

This has always been an argument for accessibility as good engineering practice. Now it's also an argument for accessibility as competitive infrastructure. As AI-powered browsing agents become a real traffic source — handling tasks, filling forms, making purchases on behalf of users — sites that are machine-readable will function in that environment. Sites that aren't will be invisible to it.

Why This Is Experimental but Not Theoretical

Lighthouse is marking the Agentic Browsing category as experimental, which is honest — the standards for the agentic web are still being written. But "experimental" here means the scoring methodology is evolving, not that the underlying problem is hypothetical.

AI agents are already browsing the web. Anthropic's computer use, OpenAI's Operator, Google's own agentic products — these systems interact with websites programmatically right now. The question isn't whether machine traffic will matter to your site. It's whether your site is built in a way that makes that traffic useful rather than frustrated.

The Lighthouse audit gives developers a concrete, CI/CD-compatible signal to work against. That's not a research project. That's a production concern.

What Your Team Should Do With This

The good news is that most of the fixes are things your development team should already be doing. Semantic HTML, proper ARIA labeling, dimensioned images, stable layouts — these are table stakes for performance and accessibility audits. Agentic readiness largely inherits from doing those things well.

The new work is WebMCP adoption: explicitly registering your site's tools and forms so agents can discover and use them reliably. And adding llms.txt to your domain root, which takes about ten minutes and signals to any AI system crawling your site exactly what it's dealing with.

For marketing teams, the strategic implication is straightforward: the next generation of search and discovery may run significantly through AI agents acting on behalf of users. A site that's invisible to those agents — because its accessibility tree is broken, its layout shifts unpredictably, or it has no machine-readable identity — is a site that loses that traffic before the conversation starts.

Build for the Web That's Coming

AI-ready marketing infrastructure isn't just a dev checklist — it's a growth strategy. If your team isn't thinking about how AI agents will interact with your site, let's change that.