AI Search for Ecommerce
For an online retailer, AI visibility runs on structured product data rather than prose. What is less widely understood is that the two big surfaces read different things. Google’s generative features work from its Search index and your Product markup. ChatGPT shopping works from a merchant product feed, and OpenAI’s published specification does not mention on-site Product schema at all. Optimising one does not deliver the other.
Rated 5.0 from 20 Google reviews · A local SEO agency based in Haywards Heath
Two surfaces, two different inputs
| Surface | What it reads | What you do about it |
|---|---|---|
| Google AI Overviews and AI Mode | The ordinary Search index, plus Product structured data and Merchant Center data | Complete, valid Product markup and an accurate feed |
| ChatGPT shopping | A merchant product feed submitted to OpenAI, per its published spec | Submit and maintain the feed; markup alone will not do it |
| Perplexity and Claude | Live retrieval of the page as it renders | Make the page readable without scripts, with price and availability in the HTML |
The middle row is the one that surprises people. OpenAI’s product feed specification is explicit about what merchants must supply and makes no reference to reading Product schema from your site. If a supplier tells you that adding markup will get your products into ChatGPT, ask them to show you where that is documented.
The nine fields OpenAI requires
From OpenAI’s published product feed spec, the required fields are: a stable item_id, title of up to 150 characters, a factual description of up to 5,000 characters, the product URL, brand, seller_name, image_url, availability, and price. Availability takes one of a fixed set of values covering in stock, out of stock, pre-order, backorder and unknown.
OpenAI also states plainly that eligibility does not guarantee display. Products have to meet its data quality standards, which in practice means the boring things: accurate prices, honest availability, real images and descriptions that describe rather than sell.
The fields most sites leave empty
- Availability, kept accurate. A feed that says in stock when it is not is worse than no feed, and it is the fastest way to be dropped from a surface.
- Brand and GTIN. Missing identifiers make a product hard to match against anything else, and matching is how a comparison gets built.
- Descriptions written as facts. Materials, dimensions, compatibility, what is in the box. Marketing adjectives are unusable to a system trying to compare two products.
- Review data, exposed properly. Aggregate ratings marked up on the page, not rendered by a script after load.
- Price, including currency, matching the page. A feed price that disagrees with the page price is a data quality failure and a legal problem.
Category pages as comparison answers
The question a shopper asks an assistant is rarely a product name. It is a comparison: which of these is best for this purpose, what is the difference between these two, what should I buy if this matters most to me. A category page that is a grid of products answers none of those. A category page that opens by answering the comparison question in three sentences, then shows the grid, answers all of them.
This does not mean long category copy. It means a short, useful answer above the products rather than a thousand words of keyword padding below them, which is a pattern that survives from an era when it worked.
Where feeds fit
If you already run a Google Merchant Center feed, you have most of the raw material for OpenAI’s spec, and the sensible order is to fix the feed once and syndicate it rather than maintain two sets of product data that will drift apart. Feed quality is now a visibility issue on more than one surface, which is a better argument for fixing it than any we had a year ago.
We have less first party ecommerce evidence than we have for local and B2B work, so this page is shorter than the others in this section and stops where our evidence does.
Frequently asked questions
Related guides
This page is one of our AI search and GEO guides. Start with the complete guide to AI search visibility or see our AI search and GEO services.
Browse all AI search and GEO guides. Would you rather we did it for you? Start with an AI visibility audit.
Sources
- OpenAI. Product feed specification. Read 3 September 2026. The nine required fields, the permitted availability values, and the statement that eligibility does not guarantee display. No reference to on-site Product schema.
- Google Search Central. Product structured data. Read 3 September 2026. The Product markup Google’s surfaces read, including price, availability and aggregate rating.
- Google Search Central. Optimizing for Generative AI Features on Google Search. Updated 10 July 2026. Google’s statement that its generative features are rooted in core ranking systems and require no special markup.
Every figure on this page was read on the source page listed above. This page is reviewed monthly and the last checked date is updated when it is.
Frequently asked questions
Do AI assistants use our Google Merchant Center feed?
Google’s surfaces can. ChatGPT uses a separate feed submitted to OpenAI under its own specification. The underlying product data is largely the same, so the practical answer is to maintain one clean source and syndicate it rather than build two.
How important is review markup?
It is one of the few places where a shopping surface can get an opinion about a product from somewhere other than you. Mark up aggregate ratings in the HTML rather than injecting them with a script, or they will not be there when the page is fetched.
Should category pages carry long copy?
No. They should carry a short answer to the comparison question at the top, above the products. Long copy pushed below the grid was written for an older set of assumptions and is read by almost nobody, human or otherwise.
What about the marketplaces we also sell on?
They corroborate, and they compete. A listing on a large marketplace is another independent record of the product existing at a price, which helps. It is also frequently the source an assistant cites instead of you, so keep your own product pages at least as informative as your marketplace listings.
Ask us how your business shows up in AI search
A friendly 15 minute call, no obligation.
or call 01444 523105