ChatGPT Product Feeds: How to Get Your Catalog Into ChatGPT

Ask ChatGPT for trail running shoes under $100 and two different things can appear. One is a set of product results that ChatGPT picks on its own. The other, below the answer, is a sponsored product ad. They look alike, but they come from two separate product feeds, and putting your catalog into one does nothing for the other.

That distinction is where most ChatGPT product-feed advice goes wrong. This guide explains which feed does what, how a merchant gets into each, what the product file needs, and how to launch catalog ads — using OpenAI's own documentation as of September 2026, linked throughout.

Which ChatGPT product feed do you need?

The rule to remember comes straight from OpenAI's product feed campaign documentation: products in an Ads Manager feed are eligible for ads only during the beta, and "will not appear in organic ChatGPT conversations." If you want both, you need both routes.

How do products get into ChatGPT's shopping results?

Organic product results are selected independently by ChatGPT and are not ads. ChatGPT weighs the shopper's request and context against structured product data from merchants and third-party providers — price, description, availability, reviews — and ranks sellers of the same item on factors including availability, price, quality and whether they are the maker or primary seller.

How you get included depends on where you sell:

• Shopify and Etsy merchants are already integrated. OpenAI's merchant page says no additional setup or application is required.
• Other merchants can apply to share a product feed directly. Applicants join a waitlist; OpenAI says it is building a self-serve merchant platform for later this year.
• No feed at all is still possible. OpenAI notes that ChatGPT also crawls sites, but a feed gives you more control over how products appear and keeps prices and stock accurate.

Two more details matter for planning. Shopping is currently live for ChatGPT users in the US only. And OpenAI says it is "moving away from a standalone Instant Checkout experience," so shoppers complete purchases on your own site or app, with no fees on purchases that start in ChatGPT.

What goes in a ChatGPT product file?

Both routes use the same basic product data, described in OpenAI's product file specification. Every row is one purchasable item or variant, and nine fields are required:

• item_id — a stable ID, never reused for a different item
• title and description — plain text, aiming for at most 150 and 5,000 characters
• url — the product page, with the variant selected where possible
• brand and seller_name — real names, not placeholders
• image_url — a direct image of this variant
• price — written as an amount and currency code, such as 79.99 USD
• availability — in_stock, out_of_stock, pre_order, backorder or unknown

Variants each get their own row, sharing a group_id. Product and image URLs must be publicly accessible. Three flags decide where an item can be used: is_eligible_search defaults to on, checkout defaults to off, and ads default to off unless your feed has an ads default configured. Feeds created inside Ads Manager, however, treat products as ads-eligible unless you mark an item false.

Most of the effort is not the format but the accuracy. A price or stock level that is out of date is worse than a missing optional field, because it sends a shopper to a page that does not match what they were shown.

How do you launch a product-feed campaign in Ads Manager?

Product-feed campaigns are built for retailers with broad or frequently changing catalogs. OpenAI says ads created from product feeds "have been among the strongest-performing ads in our program to date." Product ads show images, titles, star ratings, prices, sale prices and your brand. The setup, from OpenAI's help center:

1. Create a feed. In Ads Manager, open Tools, then Feeds, then Create Feed.
2. Import products. Upload a CSV or TXT file, point to a hosted HTTPS URL, or connect over SFTP. Items expire after two weeks, so a hosted URL or automated SFTP upload is the better choice for anything beyond a first test.
3. Validate. Processing takes minutes to hours depending on size. Check Upload History for errors such as missing columns or invalid prices.
4. Create the campaign. Choose the Product feed campaign type. Geographic targeting for these campaigns is country-level only.
5. Choose products per ad group. Filter by brand, category, price or item ID, and confirm the product count Ads Manager shows. For groupings the filters cannot express — margin tier, product line — add custom labels in each product's ads_metadata and filter on those.
6. Set the bid strategy. Maximize results is the default. It aims for the most clicks or conversions within your budget but does not guarantee a CPA, CPC or ROAS target; use a manual maximum bid if you need a ceiling.
7. Build the ad template and launch. One template per ad group; for now it only lets you choose which image field to use. Preview sample products, then submit for review.

One trap to avoid: a custom label is a way to select products, not to target people. OpenAI's example is explicit — a label of "WA" picks products carrying that label but does not limit delivery to shoppers in Washington.

How do you keep a product feed accurate?

Feeds go stale quickly, and stale data costs clicks. The developer documentation splits updates in two:

• Price, availability or title changes to existing items go through the Delta API, which accepts just the fields that changed.
• New products, new images or descriptions, or a full refresh mean re-uploading the catalog with the same identifiers.
• Sold-out items should be set to out_of_stock explicitly rather than dropped, then back to in_stock when they return.

Expect some lag in reporting. After a product-feed campaign starts delivering, products can take up to seven hours to appear in the Products tab, and a successful upload does not guarantee that every product will serve.

Should your catalog stop at ChatGPT?

ChatGPT is a large place to meet shoppers, but it is not the only place people ask what to buy. A catalog that is already clean, current and labelled is the expensive part of the work; using it in one place only leaves much of that effort idle.

You can set up and manage ChatGPT product-feed campaigns in Thrad's self-serve platform, and use the same catalog for shopping ads across Thrad's own natively integrated AI inventory. For retailers, that means one feed, one set of product groupings and one view of performance across AI placements. More on the retail use case is on our ecommerce page.

In the campaigns we run, conversation context is only part of what decides whether an ad performs. Other factors move click-through rate too, and we tune campaigns for them using what we see across our own inventory. For product ads, the practical lesson is to treat feed quality — titles that match how people ask, accurate prices, clear images — as part of the campaign, not as a one-time setup task.

A checklist before you start

1. Decide which route you need: organic results, product ads, or both.
2. If you sell on Shopify or Etsy, confirm your catalog data there is complete; otherwise apply for a direct organic feed.
3. Audit the nine required fields, variant grouping and image URLs against OpenAI's specification.
4. Plan how prices and stock stay current — a hosted URL, scheduled SFTP uploads or the Delta API.
5. Add custom labels for the product groupings you want to bid on differently.
6. Set up conversion measurement and UTMs before the first product ad serves.
7. Decide whether the same catalog should run beyond ChatGPT.

For the wider picture of buying ChatGPT ads — account setup, bidding and measurement — see our ChatGPT advertising guide. For how AI advertising fits retail more broadly, read AI advertising for ecommerce.