Shoppers have started asking AI assistants questions like "which running shoes suit flat feet" instead of typing keywords into a search box. Whether or not the assistant ever sends someone to your store, the preparation is the same: make your catalogue something a machine can read without guessing.
None of this guarantees you will be recommended. No honest vendor can promise that. What you can do is remove the reasons you would be skipped.
1. Clean, complete product data
AI systems work from structured data. Missing or vague fields are the first thing that gets you ignored. For every product, check:
- A clear, specific title, not "Item 4417".
- A real description that says what the product is, who it is for and what it is made of.
- Accurate price and currency, and a true availability status.
- Brand, identifiers such as GTIN or SKU where they exist, and good images.
- Variants (size, colour) represented properly, not as separate unrelated products.
2. Say who you are
An assistant that cites a product needs to know the merchant behind it. Make sure your business name, contact details, shipping and returns policies and legal pages are real, complete and easy to find. Vague or missing policies are a trust problem for humans and machines alike.
3. Keep structured data accurate
Product markup on your pages should match what is actually on the page. Mismatched prices or fake review counts do harm. We covered a version of this on our own site: we deliberately do not publish ratings we cannot back with real reviews.
4. Use plain language
Write the way a shopper asks. "Waterproof jacket for commuting in heavy rain" helps more than a string of internal codes. Answer the questions shoppers actually ask on the product page itself.
5. Prepare a product feed
Some AI shopping surfaces accept a product feed. Preparing one is useful even before you connect anywhere, because it forces the data work in point 1. The AI Agents add-on includes an AI Commerce section that checks catalogue readiness, flags item-level issues, and generates a private feed snapshot. See the changelog for the details.
An important caveat: preparing a feed is not the same as being listed. Merchant onboarding, product and market eligibility, and delivery to a provider are separate steps handled with that provider, and passing a readiness check does not guarantee publication.
6. Make sure your store can take the order
If an assistant hands a shopper to your store, the checkout still has to work on a phone, quickly, with the payment methods your customers use. Speed and mobile quality matter as much as they ever did.
7. Add your own agent
The other side of this is putting an assistant on your store, so shoppers who arrive can ask questions of your real catalogue. That is what the AI store agent does. We cover what to expect in the benefits of agentic commerce.
Where to start
Audit ten of your best sellers against the checklist above. If the data is thin there, it is thin everywhere. Fixing it improves your normal search, your feeds and any AI surface at the same time.