Agentic commerce readiness: a store checklist

AI shopping agents now search, compare and in some channels buy for shoppers. They do not browse your store the way a person does. They read crawled pages, product feeds, structured data and policy pages, then decide what to show. This checklist covers what any ecommerce store, on Shopify, WooCommerce, BigCommerce, Amazon or its own stack, should have in place, with each item tied to what the platforms themselves say.

New to the topic? Start with our guide to agentic commerce.

What is agentic commerce readiness?

Agentic commerce readiness means AI shopping agents can reach your store, read complete product data, understand your policies and complete or hand off a purchase. It covers five areas: crawler access, product data and feeds, structured data, readable policies, and a checkout path an agent can use.

Readiness is not a single setting. Each agent reads from different places. ChatGPT draws on product feeds and catalogs such as Shopify’s. Google uses Merchant Center feeds and pages crawled by Googlebot. Amazon’s assistant reads your Amazon listing and, for Buy for Me, can buy from brands’ own websites.

Here is the whole checklist in five steps:

The sections below take each step in turn.

Is my website ready for AI agents?

Start with access. If AI search crawlers are blocked, agents cannot read your pages. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers. Then check what you serve to agents directly, such as llms.txt, Shopify’s default agents.md file, and how your pages render.

Check your robots.txt first. Our free checker shows, crawler by crawler, whether your home, product, collection and content pages are allowed:

Our guide to AI crawlers explains which crawlers decide whether you appear in answers and which only train models.

Next, the files written for agents:

• llms.txt. A proposed summary file for AI tools. We found no statement from an AI company that its shopping answers read it, but it is cheap to keep current.
• agents.md (Shopify). Shopify manages a file called agents.md by default for every store, and serves the same content at /llms.txt.
• UCP profile. Stores that support Google’s Universal Commerce Protocol publish a profile at /.well-known/ucp, per the UCP specification.

To check or write an llms.txt for your store, on any platform:

Chrome’s Lighthouse also has an agentic browsing category. It checks six things, including llms.txt, accessibility for agents and layout stability. Chrome’s documentation calls it experimental and based on proposed standards, so read it as a hint, not a grade.

How to prepare product data for AI shopping agents?

Send complete, current product feeds to the channels that read them, and fill every relevant field. OpenAI’s feed requires nine fields and recommends refreshing at least daily. Google Merchant Center requires seven core fields plus brand. Microsoft says products with more filled-in fields rank higher.

Agents can only compare what a feed or page states. Microsoft puts it plainly: products with more filled-in fields rank higher in its results. The required fields are only the floor:

Sources: OpenAI and Google. Google also strongly recommends a GTIN where one exists, and adds: only provide one if you are sure it is correct.

Where to send feeds depends on your platform. OpenAI says Shopify and Etsy catalogs are already integrated into ChatGPT, and other merchants can apply to share a feed. On Shopify, products reach AI channels through Shopify Catalog only if they meet its requirements: among them a price above zero, a title and image, and a store that is not password protected.

Product pages are also where many stores are weakest. Adobe scored US retail pages on how much of their content machines can read, and product pages scored lowest of the three main page types:

The pages that sell are the hardest for an agent to read. That makes feeds and structured data, which hand agents the facts directly, all the more important.

Which structured data do AI shopping agents need?

Use merchant listing markup on product pages: Product with name, image and offers, and an Offer with price and currency. Add availability, shipping and return details. Google says structured data is not required for its AI search but is still a good idea, and the markup must match the page.

Google’s merchant listing documentation sets the minimum: a Product with name, image and offers, and an Offer with price and priceCurrency. The price must be greater than zero. Recommended fields include availability, shipping details, return policy, brand, GTIN and reviews.

For policies that apply to the whole store, Google recommends declaring a global return and shipping policy once, in Organization markup, and using product-level policies only to override it.

Google’s AI optimization guide is clear that structured data is not required for generative AI search, but says it is a good idea to keep using it. The same markup sits on the page for any crawler that parses it, not only Google’s.

What store policies do AI agents need to read?

Agents need your return policy, shipping costs and times, and a way to contact you, stated plainly and declared in your feeds. Google’s agent checkout requires return cost, return window and a policy link, plus at least one contact method. OpenAI’s feed has fields for returns too.

An agent that buys for someone must be able to answer “can I return it?” without guessing. Google’s Merchant Center requirements for UCP checkout spell out the minimum:

• Returns. The return cost, the return window and a link to the full policy.
• Support. At least one contact method: a URL, an email address or a phone number.
• Account standing. A Merchant Center account in good standing.
• Eligibility. Only products you opt in with the native_commerce attribute can be bought this way. Final-sale, used and refurbished items must stay out.

OpenAI’s feed carries the same facts: whether a product accepts returns, the return window in days and a link to the return policy. Final-sale items are sent with returns set to false.

Keep the numbers identical everywhere. Google asks that shipping costs in Merchant Center match your website as closely as possible. An agent that sees two different return windows may trust neither.

How do AI agents check out on your store?

There are three paths today. ChatGPT sends the shopper to your own checkout. Google’s UCP completes checkout inside AI Mode and Gemini, with you as seller of record. Your platform or payment provider may run its own agentic program. In every case, verify agents rather than blocking all bots.

The paths, as of October 15, 2026:

1. Hand-off to your checkout. ChatGPT shows products and sends the shopper to buy on your site or app. OpenAI has moved away from a standalone checkout inside ChatGPT.
2. Checkout in the channel. Google’s UCP checkout puts a Buy button on eligible listings in AI Mode and Gemini. Google’s help page lists the steps: meet the requirements, submit an interest form, complete the integration and mark eligible products. It is available to select merchants for now.
3. Your platform or provider. Shopify’s agentic storefronts are active by default for eligible stores. Payment providers and other platforms are adding support at different speeds. Our ACP vs UCP guide lists where each platform stands.

Then make sure you can tell a real agent from a scraper. The IETF’s Web Bot Auth working group is standardizing how automated clients prove who they are, including AI agents acting for users. Visa’s Trusted Agent Protocol builds on the same signature method and helps tell legitimate agents from malicious automation, using existing web infrastructure and with few visible changes to merchant sites or checkout pages. On Shopify, the help center says bot management is handled for you.

Once agents can reach you, the next question is whether they recommend you. We built SourceVision to measure that: we ask ChatGPT, Gemini and Perplexity the questions your shoppers ask, many times over, and report how often your products appear, with the range across runs. Then we draft what each product page is missing from facts you confirm, and nothing is written without your approval. See how it works.

SourceVision is in an invite-only beta.

Request an invite

Readiness is not a one-time project. Prices, stock and policies change, and so do the agents. Recheck your crawler access and feeds whenever you change your theme, platform or policies.

A simple rhythm works for most stores:

• Weekly. Confirm your feeds updated and no products were disapproved in Merchant Center or excluded from your platform’s catalog.
• After any theme, app or platform change. Rerun the crawler check and look at your llms.txt and robots.txt.
• When a policy changes. Update the policy page, your feeds and your structured data the same day, so every agent sees the same numbers.