A 10-point self-check: can your store serve AI shoppers?

Correction (2026-09-15): This article previously relied on one of our runs against Shopify's public demo store (hydrogen.shop), recorded as "stopped at a signup CAPTCHA, 20/100". We no longer use that record as evidence. The run saved no screenshots or HTML, and the same detection does not reproduce on hydrogen.shop today. On review, our detector at the time flagged a CAPTCHA whenever the page HTML merely contained strings such as "recaptcha", and this run's record consists only of such string matches. We cannot say a CAPTCHA was actually shown, and we now consider it most likely a false positive. Our 2026-09-02 note kept the number "for reference"; that was not enough. The affected sentences have been rewritten.

Most of what stops an agent looks perfectly fine to a human, so an operator cannot tell from the page alone whether their store lets agents through.

The list below follows the purchase path — entrance, forms, product data, checkout — rather than frequency. For the concept itself, see What is agentic commerce.

The 10-point self-check

1. robots.txt — check that a broad AI-bot block was not set by default by your platform or security service.
2. CAPTCHA on signup or purchase — the strongest stopping cause we see. The point is not to remove it, but to place it deliberately rather than discovering it is on every step.
3. Form labels — a machine cannot tell what a field is for unless the label is tied to it in HTML. Clicking a label should focus its input; developers can check for and aria-label.
4. Errors stated in text — a red border alone tells an agent nothing about what to fix.
5. Guest checkout — signup is the hardest stretch for agents; a guest path lets a purchase complete anyway.
6. Structured data — schema.org Product markup lets an agent read price and stock as fact rather than guess. Test with Google's Rich Results Test.
7. Price, shipping and stock in text — not buried in a banner image or "contact us", or the agent cannot compute the total.
8. Heavy JavaScript — content that only appears after interaction (infinite scroll, stock behind a tab) gets missed.
9. Unexpected popups — exit-intent modals and auto-opening chat widgets can block a walk, especially just before checkout.
10. Actually walk it — a static checker can score well while every agent stops early.

On that last point: Cloudflare's Agent Readiness Score is useful, but implementation and behavior are different things. UCP Checker's July 2026 report found 11,414 verified stores and zero with a confirmed payment implementation (State of Agentic Commerce July 2026).

Finish the checklist with a measurement

Agentrix walks an agent through your store and reports, with a score and a stop reason, where it stopped between signup and pre-checkout. Report generation is currently free. We never walk a store without consent, never bypass CAPTCHA, and never complete a purchase — when a CAPTCHA is detected the run stops, and that stop is itself the result.