article topic
sections to work through
last updated
AI platform features change quickly. This article describes what was verified on the date of the last update. Check the provider's current documentation before making an implementation decision.
A customer types into ChatGPT: "I'm looking for a trekking backpack under £60, for two- or three-day trips." ChatGPT answers with a carousel of products - images, prices and links straight through to merchants. The question is whether any of them are yours. The ChatGPT Merchants programme lets shops connect their own product catalogue to ChatGPT through a feed, much as they would with Google Shopping. This guide explains how it works, what OpenAI requires and what an e-commerce team can do right now.
In short: ChatGPT can show organic product results, and OpenAI accepts applications from merchants interested in a direct feed. Feed availability and purchase features depend on the market, the product and merchant eligibility. Preparing your data does not guarantee exposure.
What is ChatGPT Merchants?
ChatGPT Merchants is OpenAI's programme for retailers, in which a shop supplies its product catalogue through a data file (a feed). When a ChatGPT user asks a question with buying intent, the model matches products from those feeds to the query and shows them in the answer, complete with a title, price, image, rating and a link to the product page. Registration happens at chatgpt.com/merchants, and OpenAI publishes the technical feed specification in its developer documentation.
The mechanism resembles Google Shopping, with one fundamental difference: ChatGPT matches products to natural-language queries. The user does not type "trekking backpack 40l" - they describe a need: "a light backpack for weekend trips in the mountains, budget around £60, waterproofing matters." The model has to work out from the product description that this is a good answer to that need. Description quality in the feed therefore drives visibility more than it does in conventional comparison engines.
Instant Checkout and hand-off to the merchant's site
For some eligible products and merchants, ChatGPT can offer Instant Checkout. In other cases the product result leads to the merchant's site, where the customer completes the purchase. Which features apply depends on the product, the merchant, the account and the market.
Your data design should therefore support both scenarios: a result that hands off to the shop, and a checkout available for qualifying offers. On the merchant side, current prices, availability, a returns policy, merchant details and a coherent product page remain decisive either way.
Do not assume one purchase model up front. Confirm which integrations are available for your specific shop first, then prepare the data against the current specification and programme requirements.
How ChatGPT chooses which products to recommend
ChatGPT draws on several sources of product data: feeds submitted by merchants in the Merchants programme, publicly available product data (including search results and structured data on web pages) and integrations with e-commerce platforms. The model recognises buying intent in the conversation and selects products based on fit to the user's need, data quality, availability, price and trust signals such as ratings, reviews and the returns policy.
There are no paid placements in ChatGPT's shopping results - OpenAI states that the recommendations are organic. That is a window of advantage for shops that invest in data quality before the channel is commercialised with advertising.
The OpenAI product feed - field specification
The OpenAI feed specification resembles Google Shopping - deliberately, since OpenAI wants to make converting an existing feed easy - but it has its own fields and requirements. Supported formats are CSV, TSV, XML and JSON. The most important required fields are:
| Field | Requirements | Notes |
|---|---|---|
| item_id | unique ID, max 100 characters | a stable product identifier |
| title | max 150 characters | product name - conversational, not just a spec string |
| description | max 5,000 characters, plain text | the key field for matching natural-language queries |
| url | a working link (HTTP 200), HTTPS | the product page in your shop |
| brand | max 70 characters | the product's brand |
| image_url | JPEG/PNG, HTTPS | main image; additional ones in additional_image_urls |
| price | amount plus ISO 4217 currency code | e.g. 59.00 GBP; promotional prices go in sale_price |
| availability | in_stock / out_of_stock / pre_order / backorder | update at least once a day |
| seller_name, seller_url | max 70 characters / URL | merchant details shown alongside the product |
| target_countries, store_country | ISO 3166-1 alpha-2 codes | the countries you sell to |
| is_eligible_search | true / false | whether the product is eligible for search results |
| is_eligible_checkout | true / false | whether the product is eligible for checkout, where available |
Optional fields that genuinely improve the chance of a recommendation: gtin and mpn for unambiguous product identification, star_rating and review_count as trust signals, q_and_a and reviews as content the model can draw arguments from, variants (color, size, variant_dict), shipping data, and the returns policy (accepts_returns, return_deadline_in_days). For products covered by checkout, OpenAI additionally requires links to the merchant's terms and privacy policy.
The most common mistake is pasting a Google Shopping feed across unchanged. Technical descriptions in the style of "Backpack X40, 420D nylon, 38x25x60" do not answer natural-language questions. ChatGPT matches products to needs - "for two- or three-day trips", "waterproof", "lightweight" - so the description has to address those needs explicitly.
Programme availability and how to apply
OpenAI provides an application form for merchants interested in a direct feed. Access is not automatic and may depend on the market, the category, the merchant and whether the products comply with commerce policies. Status therefore has to be confirmed for a specific shop rather than inferred from a general statement about a country.
Regardless of feed access, it is worth putting public product pages, structured data, prices, availability, policies and descriptions in order. These are the controllable data foundations. Implementing them does not, however, imply feed approval or presence in ChatGPT results.
What you can do now - a shop readiness checklist
- 1Check what ChatGPT says about your shop and products today - ask 10 to 15 buying questions phrased the way your customers would describe a need, not a product name.
- 2Implement schema.org Product and Offer structured data on product pages: price, availability, ratings, GTIN. This is what AI systems draw on before any feed is connected.
- 3Prepare a feed in a format aligned to the OpenAI specification - if you have a Google Shopping feed, map the fields and fill the gaps (merchant data, policies, eligibility flags).
- 4Rewrite the descriptions of your most important products around conversational queries: which needs the product solves, who it is for and which situations it suits.
- 5Tidy up trust signals: ratings and reviews on product pages, a clear returns policy, delivery times, terms and a privacy policy served over HTTPS.
- 6Make sure AI crawlers can reach the site: check robots.txt for GPTBot and OAI-SearchBot - a blocked crawler means an invisible catalogue.
- 7Register your interest in a direct feed, and confirm the current requirements for your shop, market and category before implementing.
Why data quality is the right place to start
Structured data, complete descriptions, a current catalogue and clear policies improve the catalogue regardless of any single channel. Their effect on visibility has to be measured against a fixed set of questions and sources, without promising a position or a volume of traffic.
We treat shop readiness as a data quality and distribution project. First we establish the current state, then we improve the controllable elements, then we repeat the measurement under the same protocol.
Not sure where to start? We will assess your category fit free of charge, and scope the analysis of your feed, structured data and model answers as paid work.
Assess whether your category is suitable for AI Visibility measurement.
We assess category fit free of charge. Full visibility and competitor measurement is delivered within a Snapshot or a strategic audit.
0 PLN · Fit assessment, without a promise of a full report
Frequently asked questions
Does adding products to ChatGPT cost anything?
Taking part in ChatGPT Merchants and submitting a feed are free. OpenAI states that product recommendations are organic - there are no paid placements. The cost on the merchant side is data preparation: a feed aligned to the specification, structured data and descriptions.
Is ChatGPT Merchants available in my market?
The availability of product results, direct feeds and checkout can depend on the market, the account, the product and the merchant. A shop can tidy up its data and register interest, but this should not be treated as a guarantee of access or exposure.
How does the OpenAI feed differ from a Google Shopping feed?
An existing feed can be a starting point, but the mapping has to be checked against the current specification. OpenAI uses the is_eligible_search and is_eligible_checkout flags among others, and requires merchant data, policies, target countries and product information.
Does the purchase happen inside ChatGPT or in my shop?
Both are possible. Instant Checkout can appear for some eligible products and merchants; in other cases ChatGPT directs the user to the merchant's site. Availability has to be confirmed for a specific shop and market.
How can I check whether ChatGPT shows my products?
Ask ChatGPT buying questions that describe a need rather than a product name, and record the complete answers and sources. Analysis of the feed, structured data and a comparable set of answers is delivered within a Snapshot or an audit.