What we check
What a report covers today
Agent readiness has a lot of moving parts, and plenty of them are still draft specifications. So this page is split: what an assessment measures right now, and what we intend to add. Nothing is listed in the first section that a report will not actually deliver.
Assessed today
Every finding in these areas comes with the request we made and the response we received.
Agent discoverability
Whether an agent can find your content at all: a valid robots.txt, a reachable sitemap, canonical URLs that do not contradict each other, and Link response headers pointing at related resources.
Agent guidance
Whether you publish an llms.txt telling agents what your site is and which pages matter. A draft convention, and your report says so.
Machine-readable content
Whether your content exists in the HTML the server returns, or only appears after JavaScript runs — and whether structured data describes your business in a form software can parse.
Bot access control
AI-specific rules in robots.txt, Content Signals declaring how your content may be used, and a Web Bot Auth key directory for verifying which agents are genuine. Web Bot Auth is reported for information only and never counts against you.
Protocol discovery
Whether an agent can find the interfaces behind your site: an API catalogue, OAuth authorisation server discovery, OAuth protected resource metadata, and an auth.md describing how to authenticate. A site with no such interfaces is marked not applicable, never failed.
Foundations
HTTPS, status codes, titles and descriptions, heading structure, declared page language, image alt text, contact details, security headers, compression, caching and response time. Unglamorous, but an agent that gets a 500 or a redirect loop never reaches the interesting part.
On the roadmap — not yet assessed
These are the emerging standards we are building detection for. We list them so you know where this is heading, and so you can tell the difference between what we measure and what we merely have opinions about. What each of these is for.
Content negotiation
PlannedWhether your site can return Markdown to an agent that asks for it, instead of HTML wrapped around the content it actually wants.
Agent tool protocols
PlannedMCP server cards, A2A agent cards, Agent Skills and WebMCP tool registration — the declarations that let an agent do something on your site rather than just read it. We detect the API catalogue and OAuth side of this today; these four are not yet assessed.
DNS-level discovery
PlannedSVCB and HTTPS records under _agents that point an agent at your interfaces before it has fetched a single page.
Agentic commerce
PlannedPayment and checkout protocols that let an agent transact on a customer's behalf. Only relevant if you sell online, and the specifications are moving quickly.
What an assessment will not tell you
We would rather be useful than impressive. An assessment cannot promise that an AI assistant will cite or recommend your business — nobody can, and anyone saying otherwise is guessing. What it can tell you is whether an agent is able to reach and understand you, which is the part you control.
We also do not audit anything behind a login. The assessment reads public pages and well-known files only, which means it sees your site the way an agent sees it, and nothing more. How we handle your data.
See where your site stands
Enter a website and a business email. We confirm the inbox before anything is fetched.