audit: online store · checked: 27 October 2026 · sources: primary documents

What agent-ready means for an online store.

An agent-ready store lets an AI shopping agent find a product, read its price, stock and policies as data, buy through a route you chose, and leave a record you can audit. The sections below define each part, give a five-area audit you can run today, and say who can assess the result.

Score your store Agent readiness overview

What does agent-ready mean for a store?

A shopper's agent does not browse your home page. It looks for facts it can compare: what the product is, what it costs now, whether it is in stock, what shipping and returns apply, and how to buy. If those facts are only in images, pop-ups or scripts, the agent skips the store or guesses.

Two Origin Pi posts set the frame. The 2026 settlement is discover in AI, buy on your site: assistants do discovery and research, and the merchant keeps checkout, tax, payment and the transaction record. And structured business data is what stops agents getting your hours and prices wrong. Agent-ready ecommerce is both, applied to a catalogue.

How an AI shopping agent moves through a storeFive stages, top to bottom. Find: the agent reaches your product feed and sitemap. Read: it reads Product and Offer structured data that matches the page. Decide: it compares price, stock, shipping and returns held as data. Buy: it starts checkout on your own site through an agent checkout protocol, with a human confirm step. Prove: your log records what the agent saw and bought.1 FindProduct feed and sitemapEvery URL reachable; feed matches2 ReadProduct and Offer structured dataSame facts as the visible page3 DecidePrice, stock, shipping, returnsPolicies at stable URLs, marked up4 BuyCheckout on your siteMerchant of record; human confirm5 ProveLog of what the agent saw and boughtChannel, agent, order ID, terms shown
Figure: the five stages an agent passes through, and what the store must expose at each. The audit below follows the same order.

How do you audit a store for agents?

Check five areas, in this order. Each row names the evidence to collect and the primary document that defines it.

AreaWhat to checkPrimary source
1. Product dataOrigin Pi criteria: price, stock and an identifier in plain text. Google guidance: feed and page agree.Google: Product structured data (updated 10 December 2025)
2. Structured dataProduct with Offer: name, image, price and currency required, availability recommended. Only pages where a shopper can buy are eligible for merchant listings.Google: Merchant listing (updated 8 September 2026); schema.org/Product
3. Checkout and payment protocolsWhich agent checkout routes you support, who is merchant of record, how an agent proves a person approved the spend.ACP and AP2, as mapped in the agent-protocol map; UCP specification (its examples use version 2026-08-25), read directly (the map does not cover UCP)
4. Policies agents readReturns and shipping as text and as markup, at stable URLs.Google: Merchant return policy and shipping policy (both updated 8 September 2026)
5. Logging agent transactionsChannel, agent, order ID, the terms shown, a named owner.Origin Pi recommendation, informed by UCP's section on transaction integrity and non-repudiation: when its optional AP2 mandates extension is negotiated, businesses sign checkout terms and platforms provide mandates proving user authorization

Markup and feed work together: Google says a Merchant Center feed alongside Product markup maximizes eligibility for merchant listing experiences and helps it verify your data (Product structured data), so audit the feed route as well as the page markup. Two cautions from the sources. Google states that structured data is not required for its generative AI search features and that no special markup exists for them (Optimizing for generative AI search, updated 10 July 2026); markup earns its place through merchant listing eligibility and because it states your facts once, in a form any reader can parse. And the payment layer is still moving: the protocol map records ACP, AP2 and x402 competing, so keep the payment layer thin and own your confirm step.

Score your store

Self-assessment

Is your store ready for AI shopping agents?

Tick what is true now. Heavier items count more. The first three gaps appear under the score.

Product data and markup
Checkout and payment
Policies agents read
Logging agent transactions
  • 0 to 39%: Not yet readable. An agent would struggle to read, price or buy from this store. Start with product data and markup.
  • 40 to 74%: Partly agent-ready. Agents can read some of the store. Close the weighted gaps below, then the policy and logging gaps.
  • 75% and above: Agent-ready. The store is readable, buyable and auditable. Keep it current and re-check when a protocol version changes.

Weights are illustrative. Each item is tagged with its source or marked as an Origin Pi recommendation; none is a legal requirement. Total 33 points. Based on the primary documents in the table above, checked 27 October 2026. A guide, not an audit or legal advice. Nothing you tick leaves your browser.

Who can assess a store for agent readiness?

  • You, first. The checklist above is short and shows which of the five areas is weakest.
  • Your developer or platform partner. They can confirm markup, feeds, checkout routes and which agent protocols your platform supports.
  • An independent assessor. Origin Pi runs agent readiness assessments for businesses: how an agent reads you today, what to structure, and how to govern agent actions. See agent readiness.

For the feed and Merchant Center work that sits underneath, leapbuzz, Origin Pi's marketing consultancy brand, runs ecommerce programmes built on product feed quality.

Next step

Have a store assessed

Origin Pi

An agent readiness assessment

We review product data, markup, checkout routes, policies and logging, and give you a ranked list of fixes.

Ask for a store assessment
Questions

Common questions.

What does agent-ready mean in ecommerce?

An agent-ready store lets an AI shopping agent find a product, read its price, stock and policies as data, buy through a route you chose, and leave a record you can audit. It is five things in order: product data, structured data, checkout and payment protocols, policies, and logs.

Do I need special markup to appear in AI search?

Google says structured data is not required for its generative AI search features and that no special schema.org markup is needed (Google Search Central, updated 10 July 2026). Merchant listing data can come from Product and Offer markup, a Merchant Center feed, or both, and the same facts help any agent that reads your page.

How do I audit my store for AI agents?

Work through five areas: product data, structured data, checkout and payment protocols, the policies agents read, and the log of agent transactions. The checklist on this page scores them with weights and lists the first three gaps to fix.

Who can assess my store for AI agent readiness?

You can run the first pass with the checklist (an Origin Pi guide, not an audit). Your developer or platform partner can verify markup, feeds and checkout. Origin Pi runs agent readiness assessments, and leapbuzz, a brand of Origin Pi Pte. Ltd., covers product feeds and Merchant Center work.

What audit trail do merchants need for AI agent transactions?

This is an Origin Pi recommendation, not a legal requirement or a platform rule. We suggest at least the order ID, whether an agent or a person placed the order, which agent or platform, and a snapshot of the price, stock, policy version and terms shown at purchase time. Refunds and disputes should trace back to that record, with a named owner and a set retention period.

Sources

All read on 27 October 2026. Dates are the page's own update date where it shows one.

→

See how an agent reads your store.

A real person replies. No funnel, no sequence.

Talk to us