How the check works

A free tool to see how well your site can be read and used by AI agents. Enter a public website address to get a report showing what the scanner could access, where it found barriers and what to review next.

What happens when you start a scan

1. Find public pages. We start with the address you enter, then use links and sitemaps to discover other pages on the same website. Entering a longer URL helps guide the sample; it does not limit the scan to that one page.
2. Check reading and interaction. We inspect page content and published agent capabilities, run browser checks and a separate Lighthouse Agentic Browsing audit, and examine supported controls.
3. Attempt a test enquiry. If we find a supported enquiry form, we automatically prepare clearly marked test data and attempt one submission. This is part of the scan, not a separate add-on.
4. Explain the results. The report groups findings, shows affected pages and evidence, and separates confirmed observations from checks the scanner could not complete.

What we look at

• Access and discovery: page responses, indexing directives, crawler rules, sitemaps and supported discovery signals.
• Readable information: page text, titles, descriptions, headings, structured data and whether important content is available before JavaScript renders it.
• Navigation and controls: links, control labels, accessibility, loading, layout movement and supported form interactions.
• Agent-facing capabilities: published API, authentication, tool-discovery and commerce formats. These are optional; their usefulness depends on what your website needs to do.

How the form test works

The tool selects a suitable enquiry form automatically. It uses clearly marked test values and our dedicated reply address, tests@canagentsusemysite.com. You do not need to enter your own email address. A test enquiry may trigger a notification or another automation on the receiving website.

We check the submission attempt and the website’s response. We do not need to confirm that an email arrived in an inbox. A successful acknowledgement applies to that attempt; it does not prove every agent can submit or that the website completed every step behind the scenes.

If a security check such as Cloudflare Turnstile stops the attempt, the report identifies the observed barrier. Blocking automated enquiries may be intentional. The finding helps you decide whether that protection matches how you want agents to use your site; it is not a reason to remove security controls automatically.

Some forms cannot be tested with the supported workflow. The report distinguishes an observed block from a test that was not sent or could not be confirmed. An unclear submission outcome is not automatically retried, because the enquiry may already have been sent.

Purchases, bookings, account creation, uploads and subscriptions are excluded. We do not bypass verification challenges or log in to the website.

Which pages are checked

The scanner checks the HTML and source information of up to 20 distinct public pages. It may check fewer when fewer eligible pages are discovered, URLs resolve to the same page, requests fail, or the sample reaches its time or data limit. The report lists the pages actually checked and explains the sample’s coverage.

Browser and Lighthouse checks focus on the submitted page, with targeted workflow pages where supported. They are not repeated on every sampled page. Browser observations are brief, so slow-loading content or a scanner problem can leave some checks incomplete. Public resources advertised on other domains may also be inspected where supported.

How to read the score and findings

The readiness score summarizes completed checks of page access, readable information, navigation, control accessibility and relevant task controls before submission. The form-submission result is reported separately and does not change that numeric score. A high score can therefore appear alongside a blocked submission.

A score is shown only when all applicable weighted checks have evidence. Incomplete checks cannot improve the score. Incomplete checks are not assumed to pass or fail. The report includes the scoring breakdown and limitations.

• Needs attention: an observed issue or barrier to review, with affected pages and supporting evidence.
• What’s working: conditions the scanner confirmed. A passed check is not a guarantee that the entire website works for every agent.
• Couldn’t confirm: missing evidence because a check did not complete or could not establish an outcome. This is not a confirmed website problem.
• Optional improvements: capabilities that may help a particular use case. Missing an optional format does not automatically make a website unusable.

Keeping and sharing your report

Saved reports are private to the browser session that created them and expire after 24 hours. Reloading a report can restore it while it is still available in that session. If it cannot be reopened, you can start a new scan.

You can copy a readable summary or download the technical evidence. Reports are not published automatically. Choosing to create a public link shares a separate summary that anyone with the link can read; that link expires after seven days. Review the summary before sharing it.

What the result can tell you

Start with the findings that affect your important pages and intended agent workflows. A blocked enquiry is useful evidence about this automated attempt, not proof that every bot or agent will receive the same result. Results can change with the agent, network, login state, timing or later website updates.

The scan does not certify security or accessibility, measure traffic capacity, verify AI citations, execute discovered agent tools or prove a working checkout. Preserve intentional restrictions, review changes with your website team and test again after making them.

Check a website