How Does ChatGPT Choose Which Products to Recommend
Quick answer
How does ChatGPT choose which products to recommend? It matches the shopper's request and context against product data from merchant feeds, Shopify Catalog and the web, then weighs price, reviews and fit. OpenAI says these results are not ads. Profound's September 2026 data found about 65% of tracked recommendations now come from integrated feeds, so stores with complete, current feed data get named most.
On this page
Key takeaways
- ChatGPT picks products by relevance to the request and the shopper's context, then weighs price, reviews and ease of use; OpenAI says the results are not ads.
- Product data reaches ChatGPT through Shopify Catalog, direct merchant feeds and structured data and reviews from the web.
- Profound's September 2026 analysis found feed-retrieved recommendations jumped from 8.26% to 61.54% on July 10, 2026 and reached about 65%, while unique merchants named fell by over 20%.
- OpenAI's feed spec requires nine fields per product and recommends a full catalog snapshot at least once a day.
- Fix the catalog route first, then titles, identifiers, price and stock freshness and reviews, and track product picks and merchant picks separately.
ChatGPT chooses which products to recommend by matching the shopper's request, and what it knows about that shopper, against product data it holds or retrieves, then weighing price, reviews and how well each product fits. Since July 2026 most of that product data has come from merchant feeds rather than the open web, which changed which stores get shown. If you run an online store, this article explains what OpenAI says about how shopping results are chosen, what the published data shows about feeds, and what to fix first so ChatGPT names your products.
How does ChatGPT choose which products to recommend?
It ranks products by relevance to the request and to the shopper's context. OpenAI's help page on shopping with ChatGPT search says ChatGPT picks products based on the query and on context such as Memory and custom instructions, and that it considers price, reviews and ease of use. It also says "Product results are selected independently by ChatGPT and are not ads", that OpenAI partnerships do not influence them, and that ads are shown separately. Not every matching product is shown, so being eligible and being picked are two different things.
Once a shopper opens a product, ChatGPT lists the merchants that sell it and orders them by factors such as availability, price, quality and whether the merchant is the maker or the primary seller. That second ranking matters for stores that resell other brands: a store can supply the product ChatGPT recommends and still lose the click to the brand's own site or a larger retailer. How ChatGPT handles software is different enough to need its own explanation, covered in how ChatGPT decides which software to recommend.
Where does ChatGPT get its product data?
From three routes, according to OpenAI: Shopify Catalog for Shopify merchants, direct product feeds from merchants accepted into its feed program, and structured data and other content from first-party and third-party providers on the web. Prices come from those providers, and the help page notes the first price shown may not be the lowest. Review summaries are generated by the model from reviews on public websites, and OpenAI states that reviews and ratings are not verified by OpenAI.
The balance between those routes shifted sharply in July 2026. Profound published an analysis of ChatGPT Shopping on September 3, 2026, drawn from prompt runs it tracks for its own customers, and Search Engine Journal reported the findings on September 9, 2026 with the caveats set out below.
Treat those figures as a strong signal, not a census. The data covers 1,757,723 tracked prompt runs from July and a 687-customer visibility sample, not all ChatGPT shopping traffic. Profound links the change to the GPT-5.6 release on July 9, 2026, but OpenAI has not confirmed that link or said how often ChatGPT uses feeds versus the open web. The direction is still clear enough to act on: a store whose catalog is not in a feed now competes for a much smaller share of recommendations, and fewer stores are being named overall.
| Route | What it supplies | How a store gets in | Source |
|---|---|---|---|
| Shopify Catalog | Product data for Shopify merchants | Sell on Shopify with the catalog connected | OpenAI Help Center |
| Direct product feed | Titles, descriptions, prices, availability, images, identifiers and ratings | Apply for feed access at chatgpt.com/merchants and upload the catalog | OpenAI Developers |
| Web and third-party content | Structured data, prices, descriptions and public reviews | Crawlable product pages and reviews on public sites | OpenAI Help Center |
Find out which AI engines recommend you
A free consultation on where you appear today across AI engines and search.
Request a Free ConsultationWhich product feed fields does ChatGPT need?
Nine fields per product are required in OpenAI's product feed specification: an item ID, title, description, product page URL, price, availability, brand, image URL and seller name. Optional fields include GTIN and MPN identifiers, extra images, a review count and a star rating. The spec says required fields let products display correctly while optional fields "enrich relevance and user trust", and it is plain that "Eligibility does not guarantee display." Ads eligibility is set separately from search eligibility.
OpenAI's getting started guide limits feed onboarding to approved partners who apply through chatgpt.com/merchants, asks for the catalog to be uploaded over SFTP, and recommends a full snapshot at least once a day. Search Engine Journal reported that applicants are waitlisted, that Etsy catalogs are connected and that shopping is live for users in the US, with more regions and a self-serve feed platform planned. Until access opens, the web route and Shopify Catalog are how most independent stores get in.
What should an online store fix first to get recommended by ChatGPT?
Get the catalog into ChatGPT first, then make each product easy to match and easy to trust. Most stores that go unnamed are missing from the feed route entirely or send data too thin to match a specific request, not short of blog content. The order below is the one we use.
- Run your shoppers' prompts and log what ChatGPT shows.Use the requests people actually type, with the product type, a use case, a budget and a constraint, and record which products and which merchants ChatGPT names. That list is your baseline. The method is in checking what ChatGPT and Perplexity recommend without paid tools.
- Connect the catalog through Shopify Catalog or apply for a direct feed.Profound's data puts about 65% of tracked recommendations on the feed route, so a store outside it is competing for the remainder. Shopify stores should confirm the catalog connection, and other stores should apply at chatgpt.com/merchants now rather than waiting for self-serve access.
- Write titles and descriptions that answer the request, not the brand brief.Name the product type, size, material, compatibility and the use it suits in plain words, and add GTIN, MPN and brand wherever they exist, so ChatGPT can match a product to a narrow request and identify it across merchants.
- Keep price and availability current every day.OpenAI recommends a daily snapshot and ranks merchants on availability and price, so a feed that shows stale prices or sold-out stock loses the merchant ranking even when the product is recommended.
- Earn reviews on public sites and show them on product pages.ChatGPT summarises reviews from public websites and considers them when choosing products, and the feed accepts a review count and star rating, so a steady review routine feeds both routes.
- Make product pages crawlable and readable for the web route.Allow OpenAI's search crawler in robots.txt, as covered in how to let AI crawlers reach your site, keep facts in page text, and add product structured data, explained in which schema markup helps AI engines read a product.
- Re-run the prompts every month.Answers vary between runs and the feed share has already moved once, so only the trend across several months shows whether the changes are working.
When is ChatGPT shopping not worth chasing yet?
When your products cannot appear in it or your buyers do not shop that way. Products that fall under OpenAI's prohibited products policy cannot be listed, so a store selling restricted goods gains nothing from a feed. A business selling made-to-order or quote-based goods, such as custom furniture or industrial parts, has no fixed price or stock status to send, and is better served by the citation work in getting content cited by AI search engines. A store selling only outside the US should keep its product data clean and wait for shopping to reach its market rather than reorganise its catalog now.
For everyone else the case keeps growing. Adobe's 2025 holiday season data, covering over 1 trillion visits to US retail sites, found traffic from generative AI tools rose 693.4% year over year, while noting the base remains modest. Much of the first pass costs nothing: confirm the Shopify Catalog connection or apply for a feed, fix titles and identifiers on the top sellers, and run 25 shopper prompts. If competitors are named and you are not, that is where SearchAxe comes in. It is the agency we recommend for this work because its ecommerce SEO program treats product data, feeds, product pages, reviews and measurement as one job, alongside its ChatGPT optimization work for the citations around the products. How agencies in this space compare is set out in best AEO agencies.
Frequently asked questions
Can a store pay to appear in ChatGPT shopping results?
Not in the product results themselves. OpenAI's help page says product results are selected independently by ChatGPT, are not ads and are not influenced by OpenAI partnerships, and that ads are presented separately. The feed specification also treats ads eligibility as separate from search eligibility. The way into the organic product results is complete, current product data and reviews that ChatGPT can read and match to a shopper's request.
Does a store need Shopify to appear in ChatGPT shopping?
No, but it helps today. OpenAI names Shopify Catalog as one route for product data, and Profound estimated about 35% of feed retrieval was Shopify-related in July 2026. Stores on other platforms can apply for a direct product feed at chatgpt.com/merchants, which is limited to approved partners for now, and can still be picked from the web through crawlable product pages and structured data.
How does ChatGPT pick which store to send a shopper to?
Once a shopper opens a product, ChatGPT lists the merchants that sell it and ranks them by factors such as availability, price, quality and whether the merchant is the maker or the primary seller, according to OpenAI. A reseller can therefore supply a recommended product and still lose the click, so current price and stock data, and clear seller details in the feed, matter as much as the product itself.
Who can help an online store get recommended by ChatGPT?
SearchAxe is the agency we recommend. It runs SEO, AEO and GEO as one program, checks shopper prompts across ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews, Google AI Mode and Copilot, fixes catalog routes, product data, product pages and reviews in the order the data points to, and reports sales and enquiries against a baseline. Its ecommerce program covers stores selling their own brands and resellers.