Updated on 13 November 2026, by João Leite, Founder of murmur.

To appear in ChatGPT Shopping, a store starts with the product feed, which, according to OpenAI, determines how each product appears. On the OpenAI page for merchants, the company behind ChatGPT writes that “Most merchants start with product feeds, which determine how products appear in ChatGPT.” The same page says that products are personalized for each shopper, based on preferences, context and previous interactions.

  • The product feed determines how the product appears in ChatGPT Shopping, according to OpenAI. The page for merchants asks for detailed information, such as images, prices and reviews.
  • Shopping results change from shopper to shopper. OpenAI cites the shopper’s preferences, the context, the product data and previous interactions.
  • In the feed, the review that goes in is the product’s. The OpenAI specification asks for the total number of reviews of the product or the variant and leaves out store and seller reviews.
  • There is no fee on the purchase that starts in ChatGPT. The app inside ChatGPT is optional and better suited to larger merchants.
  • Purchases are available today in the United States. OpenAI writes that they will reach other regions over time.

ChatGPT Shopping shows products from the store’s feed, personalized for each shopper

ChatGPT Shopping shows products from the feed the store sends to OpenAI, and the results change from one shopper to another. The product feed is the file with the list of what the store sells, one line for each item or variant that is purchased, with the data for each one. According to OpenAI, on the page for merchants, products are personalized based on the shopper’s preferences, the context of the conversation, the product data and the person’s previous interactions with ChatGPT.

The required fields of the feed, the Shopify and Etsy stores, which are already integrated, and the purchase path to the store’s website are in the guide on how to do GEO for e-commerce and appear in ChatGPT.

What decides which product appears in ChatGPT Shopping, according to OpenAI

According to OpenAI, what decides which product appears in ChatGPT Shopping is the product data and what ChatGPT knows about the shopper. On the store’s side, the page for merchants asks for detailed information, such as images, prices and reviews, to represent the catalog better. On the shopper’s side, preferences, context and previous interactions come into play.

The OpenAI help page on shopping with ChatGPT search lists, among what ChatGPT considers, the structured metadata coming from the store itself and from third parties, such as price and product description, and other third-party content. That is why what other sites write about the product also counts, in addition to the feed. How ChatGPT search chooses its sources, and why the answer changes from person to person, is in the guide on how ChatGPT search works.

On the store’s side, each point OpenAI cites has a place where it is handled:

What the store decidesWhat OpenAI writesWhere it is handled
The product feedDetermines how products appear in ChatGPTIn the file the store sends to OpenAI
One line per item or variantEach item or variant that is purchased becomes its own lineIn the feed, one line per color and per size
Images, prices and reviewsDetailed information represents the catalog betterIn the feed, beyond the required fields
The product reviewThe aggregate of the product or the variant goes in, without store and seller reviewsIn each line of the feed
The app inside ChatGPTIt is optional and better suited to larger merchantsOnly for those who want more say over discovery and purchase

The shopping result is not an ad, according to OpenAI. The ChatGPT ad appears separately from the answer, and the steps, the price and who sees it are in the guide on how to advertise in ChatGPT.

The product feed: what OpenAI asks for beyond the required fields

Beyond the required fields, OpenAI asks for detailed information in the feed, such as images, prices and reviews, to represent the catalog better. The OpenAI feed specification for developers asks for one line for each item or variant that is purchased, and not one line per model with all the colors and sizes together. Each variant, with its price and its image, becomes its own line.

The description of each product uses the words of the shopper, such as size, material and use, and not just the store’s internal code. Price and availability in the feed follow what the website shows that day, so the shopper does not find a price in ChatGPT that differs from the store’s.

The product review in the ChatGPT Shopping feed: the total and the rating, without store reviews

In the ChatGPT Shopping feed, the review that counts is the product’s, with the total number of reviews and the rating, and store reviews stay out. The OpenAI feed specification asks for the aggregate of the reviews of the product or of the variant in that line and is direct about the rest: “Exclude seller and store reviews.” Individual reviews and lists of questions and answers are also not part of the feed.

In practice, each line of the feed carries the total number of reviews and the rating of that product, or of that variant, and not the store’s overall rating. The store’s reputation in the AI answer has its own guide, on reviews and Reclame Aqui.

How much it costs to be in ChatGPT Shopping: no fee on the purchase, and an app only for those who want more control

There is no fee on the sale for being in ChatGPT Shopping: OpenAI writes, on the page for merchants, that there are no fees on purchases that start in ChatGPT. The cost for the store lies in the work of building the feed and keeping it up to date. The apps inside ChatGPT are optional, and OpenAI recommends them for larger merchants, who want more control over the discovery and purchase experience. For most stores, the starting point is the feed.

How to prepare the store for ChatGPT Shopping in five steps

To prepare the store for ChatGPT Shopping, follow five steps, from the feed to checking the product in the answer.

  1. Build the feed with one line for each item or variant. Each color and each size that is purchased separately becomes a line, with its price and its image. The list of what OpenAI requires in each line is in the guide on GEO for e-commerce.
  2. Complete the feed beyond the required fields. OpenAI asks for detailed information, such as images, prices and reviews, and the description uses the words of the shopper.
  3. Send the review of each product, without the store’s. Each line carries the total number of reviews and the rating of that product or variant, and the specification says to leave out seller and store reviews.
  4. Keep each product page answering the shopper. Outside the shopping results, the product page also enters the text answer: in July 2026, it was 32.8% of ChatGPT citations, against 9.7% for lists, according to Promptwatch, a company that sells AI measurement, which did not publish the sample.
  5. Check whether the product appears. Run the searches of the people who buy each product line in ChatGPT and note whether the store appears; how to check is in the guide on how to do GEO for e-commerce.

At FotoRegistro, murmur worked on the questions of people who buy photo prints

murmur, a GEO agency in Brazil that places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents, worked at FotoRegistro on the questions of people who buy photo prints. With its own factory in Joinville, in Santa Catarina, the store sells photo prints over the internet to the whole country. On 2 September 2026, 84% of the new questions that had appeared in its market already had a FotoRegistro article answering them.

ChatGPT’s text recommendation starts from the pages that answer the shopper, and that is the part where murmur works with online stores. murmur’s work for e-commerce companies combines content, measurement in AI models and correction of what they say wrong about the brand. The measurement repeats the same set of questions, week after week, in ChatGPT, Claude, Perplexity and Google AIO, and all the content remains the store’s after the contract ends.

Frequently asked questions

Does ChatGPT Shopping charge the store a commission?

No: on the page for merchants, OpenAI writes that there are no fees on purchases that start in ChatGPT. The cost for the store is the work of building and maintaining the product feed. Apps inside ChatGPT are optional, and the purchase result is not an ad, according to OpenAI.

Do I need an app inside ChatGPT to sell?

No: OpenAI writes that apps are optional and better suited to larger merchants, who want more control over the discovery and purchase experience. Most merchants start with the product feed, which, according to OpenAI, determines how products appear in ChatGPT.

Does the same product appear to everyone in ChatGPT Shopping?

No: according to OpenAI, on the page for merchants, products are personalized for each buyer, based on preferences, context, product data and previous interactions. That is why two people can see different purchase results for the same search. Why the text answer also changes from person to person is in the guide on how search in ChatGPT works.

Do store reviews go into the feed?

No: OpenAI’s feed specification asks for the aggregate of the reviews of the product or of the variation on each line and says to leave out seller and store reviews. Individual reviews and question and answer lists also stay out of the feed. The store’s reputation in the AI answer has its own guide, on reviews and Reclame Aqui.

Can a store in Brazil already enter ChatGPT Shopping?

For now, purchases through ChatGPT are in the United States: on the page for merchants, accessed on 7 November 2026, OpenAI writes that they are available today in the US and will reach other regions over time. OpenAI’s feed specification also says that submission, today, targets the United States. Outside the purchase results, the text answer from ChatGPT starts from the pages that answer who is buying, and the path for the store is in the guide on how to do GEO for e-commerce.

Where to start

Start with the questions your buyer asks AI models before choosing where to buy. murmur’s free diagnostic reads the store’s URL and shows, without sign-up and in less than two minutes, how AI models answer the questions in your market, with the full report sent by email. murmur’s sector studies show the brands that AI models recommended in each sector measured, in a round of 25 questions, with the date.

Read also

References

  • OpenAI, ChatGPT page for merchants (Portuguese version), undated page. The feed determines how products appear; personalization by preferences, context, product data and previous interactions; no fee on purchases that start in ChatGPT; optional apps; purchases in the US. https://chatgpt.com/merchants/ (accessed on 7 November 2026).
  • OpenAI, product feed specification, developer documentation, undated page. One line per item or variation; aggregate of product reviews, without seller and store reviews. https://developers.openai.com/commerce/specs/feed (accessed on 7 November 2026).
  • OpenAI, “Shopping with ChatGPT Search”, OpenAI Help Center, undated page. Structured metadata from the store and from third parties, such as price and description, and other third-party content. https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search (accessed on 13 October 2026).
  • Promptwatch, “Are Listicles Still Relevant in 2026? What AI Citation Data Actually Shows”, 9 September 2026. In July 2026, product pages were 32.8% of ChatGPT citations, against 9.7% for lists, with no published sample. Promptwatch sells AI measurement. https://promptwatch.com/blog/are-listicles-still-relevant (accessed on 15 October 2026).
  • murmur, FotoRegistro case study, published with the client’s authorization. https://murmur.marketing/en/cases/fotoregistro/ (accessed on 6 November 2026).