GeoChrome's own auditing tool now checks whether your site has an llms.txt file. In the same month, Google told site owners they don't need one. Both statements are true, and the gap between them is worth understanding before you spend an afternoon on either one.
What did Chrome's Lighthouse update actually add?
Lighthouse 13.3, released in May 2026, added a new Agentic Browsing category that scores how well a page works for AI agents rather than human visitors. It doesn't produce the familiar 0 to 100 score you're used to. Instead it reports a pass or fail ratio across four checks, and one of them is your llms.txt file.
The category sits alongside Lighthouse's existing Performance, Accessibility, Best Practices, and SEO scores, but Google built it differently on purpose. The underlying standards, llms.txt and WebMCP among them, are still young, so a single weighted number would have implied more certainty than the evidence supports. Instead you get a pass count and a list of what failed (DebugBear's rundown of the release).
Does Lighthouse checking llms.txt mean Google uses it to rank you?
No. Days before this Lighthouse update shipped, Google published new AI search guidance stating plainly that llms.txt files aren't needed for generative AI visibility. Search Engine Land pointed out the irony directly: Chrome's own audit tool now flags a file Google says has no effect on rankings (Search Engine Land, May 2026).
Google's John Mueller addressed the contradiction on Bluesky when a reporter asked why Google would publish llms.txt files itself while saying they don't matter for search. His answer drew a line between two different jobs: "The short answer is that it's not done for search. There's more to websites than just SEO." He separated discovery, meaning how a search engine finds your pages in the first place, from functionality, meaning what happens once an agent has already arrived and is trying to complete a task. llms.txt lives in the second bucket, not the first.
What exactly does the llms.txt check look for?
Lighthouse doesn't just look for the file's existence. If you have one, it checks whether it's missing an H1 heading, is too short to be useful, or contains no links at all. If the file returns a 404, Lighthouse marks that check "not applicable" instead of failing you, since the file is still optional. It's a quality check on top of a presence check, not a pass/fail gate every site has to clear.
What else does the Agentic Browsing category check?
Three other checks round out the category, and none of them are about llms.txt:
- Accessibility tree structure. Agents read this simplified version of your page instead of parsing full HTML, so a broken or missing tree slows them down the same way it slows down a screen reader.
- Layout stability (CLS). An agent that screenshots your page mid-load can get confused by content jumping around, the same shift that already hurts your Core Web Vitals score for human visitors.
- WebMCP validation. WebMCP lets a page expose specific actions, like "add to cart" or "submit this form," as machine-callable functions instead of forcing an agent to guess at buttons and fields. Lighthouse checks whether any annotated forms match the expected schema.
Google Cloud's AI engineering director, Addy Osmani, described the broader thinking behind this shift as agents working with smaller context windows than a human has patience for: clean semantic structure, token-efficient content, and markdown delivery all help an agent get through a page without missing what matters.
Should you add an llms.txt file because of this update?
If you don't have one, adding it costs almost nothing and closes a checkbox in a tool more people are starting to run. Just don't expect a ranking or citation bump from it. Put your real effort where it actually moves the needle: crawler access for AI bots, structured data that names what you are, content that answers real questions in plain language, and off-site mentions in places AI engines already trust. Those four categories decide whether you get cited. llms.txt is housekeeping, not a growth lever, and GeoRoar's own rubric weights it accordingly at 4 of 100 points.
How do you check your own Agentic Browsing score?
Chrome 150 and later ships the category automatically in DevTools. Open the Lighthouse tab, run
a report, and it appears at the top with everything else. There's no flag to enable. If you'd
rather test from a terminal, install the CLI with npm install -g lighthouse@latest and
run lighthouse --view yoursite.com. PageSpeed Insights still runs an older Lighthouse
version, so DevTools or the CLI is the faster route to see where you stand right now.
GeoRoar's free audit scores all of that in one pass, including whether your llms.txt file exists and is formatted well, weighted honestly against the signals that actually decide whether AI engines cite you.
Good enough to get cited?
Here's a sample GeoRoar scorecard. Yours takes about 30 seconds.
This is a sample. Your real scorecard is free, no sign-up.
Run your free GEO audit