How ChatGPT shopping works for brands
How products reach ChatGPT shopping results as of October 2026: feeds, Shopify Catalog, crawling and ads, and what API research can and cannot tell a brand.
As of October 2026, a product can reach ChatGPT shopping results through Shopify Catalog, an approved direct feed, or open web crawling, and OpenAI says the results are not ads. What OpenAI does not publish is how much each input counts. Our own data comes from an API, so it cannot see the product carousel in the app.
How can a product appear in ChatGPT shopping results?
OpenAI's and Shopify's pages describe two routes for merchant data and one open route. Ads are a fourth thing and are kept apart.
| Route | Who it is for | What OpenAI or Shopify says |
|---|---|---|
| Shopify Catalog | Shopify stores selling to U.S. customers | Product data is already integrated through Shopify Catalog, and OpenAI says no additional work is needed from the merchant. |
| Direct product feed | Merchants who apply and are approved | OpenAI's developer pages say feed onboarding is open to approved partners. Its merchant page describes shopping as live in the U.S. and a self-serve platform as planned. |
| Web crawling | Anyone whose pages OpenAI's search crawler can read | OpenAI says a feed is not required, but gives more control. Sites that opt out of its search crawler will not be shown in ChatGPT search answers. |
| Ads | Advertisers | Paid placement shown below a response. Separate from product results. |
The feed mechanics, field by field, are in product feeds for AI shopping. AI crawlers and robots.txt have their own post.
What does OpenAI say decides which products show?
OpenAI's help article says a product appears in the carousel when ChatGPT judges it relevant to the shopper's intent, using the question and its context, which can include Memory and Custom instructions. For something like a dog costume, it says ChatGPT may weigh available options, price, reviews and ease of use. If the shopper states a budget, price gets more weight; if not, other aspects may.
When ChatGPT picks products, it considers three groups of inputs, according to the same article: structured metadata from first-party and third-party providers (price and description are the examples given) and other third-party content; the model's own earlier responses before new search results; and OpenAI's safety standards and product policies.
Several details in the article are useful to a brand:
- Titles and descriptions may be rewritten. ChatGPT may generate simpler titles and descriptions from provider data, because merchants often use different wording for the same product.
- Labels and review summaries are generated. Labels such as "Budget-friendly" or "Most popular" are produced by ChatGPT and are not verified. Review summaries are model-generated from reviews on public websites, and OpenAI says it does not verify reviews and ratings.
- Merchants are ranked. When a shopper selects a product, ChatGPT may list merchants. OpenAI says they are ranked on factors such as availability, price, quality, and whether the merchant is the maker or primary seller of the item.
- Not everything is shown. OpenAI says not all available products will necessarily appear.
- Price lag exists. OpenAI says there can be a delay between a merchant changing a price or shipping terms and the change showing in ChatGPT.
The maker-or-primary-seller factor is worth a second look for direct-to-consumer brands. If you make what you sell, that is a stated ranking factor for the merchant list. OpenAI does not say how heavily it counts.
Is ChatGPT shopping paid or organic?
OpenAI draws a hard line between the two, and the two systems are described differently.
| Product results | Ads | |
|---|---|---|
| Who chooses | ChatGPT, based on relevance to the shopper's intent | An ads system that weighs relevance and advertiser bids |
| Paid? | OpenAI says results are not ads and not influenced by partnerships | Yes, a paid placement |
| Where | Product carousel and merchant lists | Below a response, separate from product results |
| Plans | Not stated by plan in the article we read | OpenAI says ads may appear for Free and Go plans, and not for Plus, Pro, Business, Enterprise or Edu |
| Landing page checks | Not applicable | OAI-AdsBot may visit pages submitted as ads |
OpenAI says ad testing began in the United States on February 9, 2026 and is rolling out gradually to eligible users in select regions. It says ads do not influence ChatGPT's answers, and that advertisers cannot shape, rank or change responses. For a feed, the practical consequence is that search eligibility and ads eligibility are separate flags. Switching one on does not switch on the other, and checkout needs its own integration.
Where does the purchase happen?
On the merchant's site or app, as of October 2026. OpenAI's merchant page says it is moving away from a standalone Instant Checkout experience, and that purchases are completed on merchant-owned websites or apps. It also says there are no fees on purchases that start in ChatGPT. Shopify's help center describes the ChatGPT channel as a discovery-focused referrer: customers finish the purchase on your store's checkout, inside a ChatGPT in-app browser or in a new tab on the web, so your branding, payment methods and checkout customizations apply.
OpenAI's launch post for shopping research said buyers would in future be able to purchase directly for merchants in Instant Checkout. We found no current OpenAI page that describes that plan, so we do not treat it as a live feature.
That launch post is still a useful primary source on how the research mode works. OpenAI described it as an experience that asks clarifying questions and researches across the internet. It said results are organic and based on publicly available retail sites, and it pointed to an allowlisting process for merchants who want to be sure they can appear. We could not confirm from OpenAI's current help pages whether that process is unchanged.
What can API research tell a brand about the app?
Some, and not the part that is easiest to see. Our October 2026 numbers come from OpenAI's API with web search (gpt-6-luna), with 800 answers across 160 categories in the United States, Mexico and Canada. Each request carries no account, search history or memory, and the web search location is set to the market's country (see the methodology). The ChatGPT app can use Memory, custom instructions, a user's location and a product interface we do not call.
| Part of the experience | In the app, per OpenAI | In our API sample |
|---|---|---|
| Answer text naming brands | Yes | Recorded |
| Linked source pages | Yes | Recorded |
| Product carousel and merchant list | Yes | Not recorded |
| Ads below the response | In test, Free and Go plans | Not recorded |
| Memory and custom instructions | Used when relevant | Not used |
What we can say with a denominator:
- In 99.6% of the 800 OpenAI answers, at least one page was linked. An answer linked 2.2 pages on average, drawn from about 2.0 distinct domains.
- Excluding the alternative-to-a-brand question (the October prompt that asked for an alternative to a named brand), an OpenAI answer recommended 2.2 brands on average (median 2) across 640 answers, and 4.8% recommended none.
- Among all 800 answers, 46.6% linked a domain that matches a recommended brand's own name, 17.6% linked a domain on our list of 49 major retailers, and 1.3% linked a forum or video site.
The own-site rule is strict: a domain whose first label equals a recommended brand's name, so it misses country domains, parent companies and other store names. The retailer list is a hand-made list of 49 domains, and the forum and video list is short too. Each share counts answers that link at least one matching domain, so these are presence counts, not influence. Because the retailer list is short, the shares are not a like-for-like ranking of source types, and neither says what the product carousel in the app is built from. The distinction between what an API records and what an app shows is the subject of API answers versus app answers, and how assistants pick products in general is in how AI assistants choose products.
How can you check your own products in ChatGPT?
- Write five questions a shopper would ask. A best-overall question, a budget question with a price ceiling, a premium question, a specific-use question, and a plain buying-advice question that names no brand.
- Start clean. Use a temporary chat and, if OpenAI offers an Unpersonalized choice, select it before you start so saved memories are not used. OpenAI's shopping article lists Memory and Custom instructions as context ChatGPT considers, so a normal new conversation may still draw on them. Turn custom instructions off or note them in your record.
- Record what you see. For each answer, note which products are shown, the price on the first listing, the merchant list order, and any label such as "Best price".
- Compare with your own data. Check the price, availability and title shown against your product page and your feed. OpenAI says the first merchant's price shown may not be the lowest.
- Report errors. OpenAI says you can report a product that has incorrect information from the menu on the product.
- Repeat on different days. Answers vary. Our free manual method explains how many repeats to run per question (four) and how to record them.
If you sell on Shopify, OpenAI says the data flows through Shopify Catalog. In your Shopify admin, check the settings of the ChatGPT channel to confirm that access has not been turned off (the exact menu names can change).
What is still unknown?
- How much each input counts. OpenAI lists inputs. It does not give weights, and we did not test them.
- Whether a feed changes ranking. OpenAI says a feed gives more control over how your products appear, and does not say it ranks higher.
- Memory effects. Personalization changes what a given shopper sees, and our samples carry none.
- Regional rules. OpenAI's merchant page describes shopping as live in the U.S. Check the current pages for your market.
- In-chat purchase. OpenAI's merchant page says it is moving away from a standalone Instant Checkout. Whether any in-chat purchase remains is not stated.
Frequently asked questions
Does blocking GPTBot remove my products from ChatGPT shopping?
Not by itself, according to OpenAI: disallowing GPTBot signals that content should not be used for training, while opting out of OAI-SearchBot keeps a site out of ChatGPT search answers. OpenAI says each crawler setting is independent of the others, and does not say how either one affects shopping results specifically. Shopify adds that data sent through Shopify Catalog does not depend on robots.txt. Check each layer separately.
How long does it take for a change to show up?
OpenAI gives no figure for shopping results, and says there may be a delay after price or shipping changes. For search crawling, it says changes to robots.txt can take about 24 hours to be reflected. Shopify says turning off Catalog access can take up to 7 days.
How we write: drafted with AI tools from Dopbot Index data and public sources; every number and outside fact is traced to its source in a claims record and checked before publishing. More about how we write; corrections to hello@dopbot.com.