AI agent readiness scan
Can an AI agent use your website?
Paste your address and we’ll check what an agent meets when it arrives. Whether it gets in, whether there’s anything to read before scripts run, whether the page explains itself and whether the forms can be operated.
Every finding is a measurement with the number behind it, scored by fixed rules. It’s free and needs no email.
What it measures
Four questions an agent’s visit answers.
Our own site · version2.hrScanned Sep 23, 2026
97of 100
Agents can work with this site.
- 01
Can an agent get in?
The server’s answer, the robots rules, and what happens at an address that doesn’t exist.
6 of 6 passed
- The page answers an automated visitorPassed
- robots.txt lets agents inPassed
- Served over HTTPSPassed
- A missing page says it’s missingPassed
- A sitemap lists the pagesPassed
- The address resolves without a detourPassed
- 02
Is there something to read?
What’s in the first response, before any script runs, and how much of it there is.
4 of 5 passed
- The content is in the HTMLPassed
- The page is light enough to read wholePassed
- The page says what language it’s inPassed
- An llms.txt maps the sitePassed
- The page is available as MarkdownFailed
- 03
Does the page explain itself?
The title, the headings, the landmarks and the structured data an agent uses to work out what it’s looking at.
7 of 7 passed
- A title and a descriptionPassed
- Headings form an outlinePassed
- Landmarks mark the parts of the pagePassed
- Structured data describes the businessPassed
- Images carry a descriptionPassed
- A canonical addressPassed
- Language versions are declared correctlyPassed
- 04
Can an agent operate it?
Whether links, buttons and form fields carry a name, and whether a declared API or MCP server is valid.
2 of 2 passed
- Form fields have labelsDoesn’t apply
- Links say where they goPassed
- Buttons say what they doPassed
- A declared API or MCP server is validDoesn’t apply
What to fix, in order
- 01
The page is available as Markdown Failed
Asked for text/markdown and got text/html.
When a request’s Accept header asks for text/markdown, answer with a Markdown version of the page and add Vary: Accept.
What this can’t see
The scan reads one public page at a single moment.
- Pages behind a login, and anything that happens after a form is sent.
- How your site answers a visitor from another country.
- A bot defence that lets us through and turns a different agent away, or the other way round.
- Whether an agent can finish a task such as a booking. That takes a person watching one try.
- A temporary outage on your side reads as a failure here too, so a second scan is worth a minute.
Questions
What people ask about agent readiness.
What is an AI agent, and why would one visit my website?
An agent is an AI system that does a task for a person, such as finding a supplier, comparing prices or booking a table. To do that it fetches web pages the way a browser does, except that it reads the HTML and usually doesn’t run scripts or look at the design. If your site can’t be read that way, the agent works from somebody else’s site instead.
How is the score worked out?
Each check is a fixed rule with a weight. A check that decides whether an agent gets in at all counts three times as much as a courtesy such as llms.txt. A check that doesn’t apply to your site, such as language versions on a single-language site, leaves the sum entirely. The score is the weighted share of the checks that do apply.
Does a model write the findings?
No. Every sentence in the report is a fixed template with your measured numbers filled in, which is why the same site gets the same report twice in a row. The free website analysis is where a person reads your site and writes about it.
Do you store anything about me?
We store the report under the domain that was scanned, so a popular site is scanned once and a second scan can show what changed. We don’t ask who you are and the report holds nothing about the person who asked for it.
My site blocks AI crawlers on purpose. Is that wrong?
It isn’t. Keeping training crawlers out is a choice about your content, and the scan doesn’t mark it down. What it does flag is a robots.txt that also shuts out the agents fetching a page because a customer asked about you, since that turns a visitor away.
Will the scan slow my site down or show up in my analytics?
It makes about a dozen small requests over a few seconds and identifies itself as Version2AgentScan, so you can find it in your server log. It doesn’t run scripts, so script-based analytics never see it.