ChatGPT shopping: what makes your products discoverable

Correction, October 4, 2026: An earlier version said assistants do not browse your website and implied that a product cannot be mentioned without a feed. Those claims were too broad. OpenAI documents both merchant feeds and public retail pages as sources for shopping research. We have corrected the explanation and removed unsupported guarantees about review counts.

ChatGPT shopping can put a product in front of a buyer before that person reaches the merchant's store. For a small retailer, the useful response is to make the product easy to identify and compare, then make sure the destination page supports the same facts.

A feed is one route for supplying those facts. It is not the only route, and submitting one does not guarantee that your product will be recommended.

What the current documentation establishes

OpenAI's shopping research guide describes a system that can use merchant product data, publicly available product information and other retail sources. It explicitly says shopping research reads retail pages and cites sources. Sites that block access may be skipped or represented through other sources.

OpenAI's merchant page says Shopify and Etsy catalogs are already integrated, so individual sellers there do not need an additional feed application. Other merchants can apply. The page currently describes shopping availability for ChatGPT users in the United States, with broader access planned. Check this scope before designing a campaign around it.

These are documented product capabilities and access rules, checked October 4. They do not establish how much traffic your store will receive.

Check one product from discovery to checkout

Pick a product with meaningful revenue and a clear customer use case. Compare the public product page, its feed record if you have one, and the information shown in a dated shopping result.

Check	What to compare
Identity	Product name, brand, variant, size and color
Price	Displayed price, active discount and conditions
Availability	Whether that exact variant can be purchased
Buyer constraints	Dimensions, materials, compatibility or fit
Handoff	Landing page, delivery terms and return policy

Record the question, date and result you saw. If the wrong color or an old price appears, locate the likely underlying record and correct it. A single answer is a useful diagnostic sample, not a complete visibility report.

For a deeper catalog checklist, use our product discovery owner's guide. Keep improving the existing product information while waiting for any integration approval.

Keep reviews honest and useful

Reviews can help a buyer assess an item, but a particular review count does not guarantee an assistant will choose it. Remove the idea that two hundred reviews will always beat fifteen.

Our recommendation is to request honest feedback consistently, without selecting only customers expected to be positive. Keep product reviews attached to the right variant where possible. Do not invent testimonials or present staff-written praise as independent customer experience.

The useful question is whether the evidence answers a buyer's uncertainty. A detailed comment about a bag's capacity can explain more than an unsupported claim that it is perfect for everyone.

Measure what happens after the mention

Keep observed appearances, recorded visits and completed orders separate. A mention without a click is not a session, and a session is not a sale. Our AI traffic measurement guide explains the reporting distinction.

OpenAI's shopping help page also warns that generated virtual try-on images can be inaccurate and do not guarantee fit. Apparel sellers should keep measurements and return conditions easy to find when shoppers arrive with expectations created elsewhere.

Your move: audit one valuable product's data and destination page, record the discrepancies, then fix the source information you control. Expand only after that handoff is reliable.