GeoEvery few weeks a new tool offers to scan your site for "AI agent readiness." Before you pay for one, it helps to know what it actually measures, because it is a different thing from getting cited by ChatGPT. This post covers what the scan checks, what the traffic data says, and where a small business should spend its first hour.
What is an AI agent readiness scan?
An AI agent readiness scan tests whether software agents, such as agentic browsers, can use your website to complete a task. It checks that forms and buttons are labeled, that the page structure is valid, and that optional files like llms.txt exist. It does not tell you whether ChatGPT or Perplexity will cite you.
That difference matters. A citation happens when an engine reads your page and quotes it. An agent action happens when software clicks, types, and submits on a person's behalf.
Curious how your own site scores with AI search?
Run your free GEO auditWhat does Chrome's agentic browsing audit check?
Lighthouse 13.3 added an "Agentic Browsing" category with four checks: whether the accessibility tree is well formed, layout stability, WebMCP validation, and llms.txt. DebugBear reported the release on May 10, 2026. The category is still marked as under development, and it does not give a 0 to 100 score.
The iO tech hub write-up calls it "more of a diagnostic than a grade," and DebugBear notes that a site does not fail just because it has not adopted new AI features. DebugBear also says llms.txt "is not widely used by AI tools currently." We covered the Lighthouse details in our earlier post on the agentic browsing audit, so this post is the buyer's view.
How much traffic do AI agents actually send?
Not much to local businesses yet. HUMAN Security measured agentic traffic growing 5.26% month over month in April 2026, and 98% of it landed on media, ecommerce, and travel sites. Two agentic browsers, Comet and Atlas, produced about 70% of the volume.
Of the pages those agents touched, 69.57% were product and search routes, and only 3.16% were checkout and payment pages (HUMAN Security, April 2026). So the early agent economy is people comparing products and travel options. A dentist's contact form barely appears in that data.
The same company's annual report, as summarized by Implicator, put agentic browser growth at 7,851% year over year in 2025. That number started from almost nothing in 2024, and Implicator's wording is ambiguous about whether it covers agents or agentic browsers, so check HUMAN's original report before quoting it.
Which agent readiness checks are worth doing first?
Label every field and button on your booking or contact form, keep the page HTML clean, and make sure nothing important is hidden from the accessibility tree. These are old accessibility basics. They help screen reader users today and agents later, so the work pays off twice.
WebMCP and llms.txt come after that. WebMCP is in an origin trial on Chrome 150 and later, so there is little reason for a small site to build against it yet. An llms.txt file takes minutes to add and does no harm, but we have never claimed it moves AI citations, and neither does the evidence so far.
What should come before agent readiness?
Citation basics: letting AI crawlers in through robots.txt, serving content without JavaScript, adding schema that names your business type, and writing answers an engine can quote. If an engine cannot read or trust your site, an agent will not rescue it.
For a local business we would go in this order. First, confirm crawlers can read your pages. Second, fix schema and answerable content. Third, run an agent check on the one form that matters, usually booking. If a scan puts "agent readiness" ahead of those three, it is selling you the new thing instead of the useful thing.
Where do GeoRoar and the scan fit together?
GeoRoar scores your site on a fixed 100 point rubric covering machine access, structured data, answerability, entity clarity, off-site corroboration, and trust. The audit is free. The Fix Pack is $99, one time, and generates the corrected files for you. Monitor is $19 a month if you want to keep watching as your site changes.
We do not score agent actions today. When the standards settle, we will add it to the rubric and say so here. Until then, the honest advice is to spend your first hour on the basics above.
Can you score better than this?
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