A human shopper works around ambiguity without thinking. An agent doesn’t: it gives up, or it picks a competitor whose information it can use immediately. Your analytics only measure humans.
Does the agent understand what you sell, and can it find a product that meets price and availability constraints?
Can it tell two items apart, understand their differences, and recommend one?
Can it pick a variant, fill a cart, understand shipping, and reach the checkout?
Your store and an email. Nothing to configure, no account to create.
An AI agent looks for a product, compares two, then tries to reach checkout.
Every stage we credit is tied to a page the browser actually loaded.
What blocks, why, and what to fix. In minutes.
An agent claiming it reached checkout is not proof that it did. Every stage it declares is cross-checked against what the browser actually recorded — and loading an address is not evidence of what happened there. When the evidence is missing, the report says so instead of selling you a success.
A claim is marked confirmed only for what the observation actually shows — the page it was seen on, and no wider. Anything a finding adds beyond that is published separately, as an observation to confirm, and counts for nothing.
Wording taken from the reports themselves.