ChatGPT product feed for Shopify: the required fields
The ChatGPT product feed is the list of products that a store supplies to ChatGPT for shopping answers. OpenAI's specification, as we read it on 14 November 2026, lists these fields as required for every item: item_id, title, description, url, brand, seller_name, image_url, availability, price, target_countries and store_country. A Shopify store does not build this feed. Shopify sends eligible products to ChatGPT through Shopify Catalog, using the product details in your admin, so the feed file matters mainly for stores that are not on Shopify or that send OpenAI a feed directly.
Checked on against OpenAI's product feed specification, and on 13 November 2026 against Shopify's Help Center. By Edesigno. Sources are linked in each section.
What the ChatGPT product feed is
A product feed is a structured list of everything a store sells: one row for each item, with the same named fields in every row. OpenAI's Help Center says ChatGPT picks products using structured product data, such as price and description, from data providers and from stores themselves, as Search Engine Journal reported in September 2026.
Feeds now supply most of ChatGPT's product picks in one vendor's tracking. Profound runs its customers' prompts on ChatGPT every day. It reported that the share of ChatGPT Shopping recommendations it classed as coming from feeds rose from 8.26% to 61.54% on 10 July 2026, across 1,757,723 tracked prompt runs that month, and stood at about 65% on 3 September 2026.
Profound's figures describe its own customers' prompt runs, not every shopper's session, and they show where products were retrieved from, not how they were ranked.
Source: Profound, reported by Search Engine Journal, 9 September 2026.
The required fields
OpenAI's specification marks these fields as required, including two country fields. One row describes one item you can buy, or one variant of it.
| Field | What it must contain | Example |
|---|---|---|
| item_id | A stable ID, unique for each item or variant in your feed. Never reuse it for a different item. | MUG-350-BLUE |
| title | The product name, including the selected variant when relevant. Aim for 150 characters at most. | Blue ceramic mug, 350 mL |
| description | A factual description of this item. Aim for 5,000 characters at most. | Dishwasher-safe glazed ceramic mug with a handle. |
| url | The product page for the item, with the variant selected when possible. Keep the address stable. | https://example.com/products/mug-blue |
| brand | The product brand, as shown on the product page. | Northline |
| seller_name | The name of the seller supplying this offer. | Northline Home |
| image_url | The main product image, showing this variant, as a direct link to a JPEG or PNG file. | https://example.com/images/mug-blue.jpg |
| availability | One of in_stock, out_of_stock, pre_order, backorder or unknown. A row with a missing or unrecognized value is rejected. | in_stock |
| price | The regular price: an amount, a space and a three-letter currency code. | 18.00 USD |
| target_countries | The countries the item is offered in, as uppercase two-letter country codes. The specification says standard uploads currently use the US whatever this field says. | ["US"] |
| store_country | The country of the seller's store. The specification says standard uploads currently use the US whatever this field says. | US |
Source: OpenAI commerce documentation, Products feed specification. Some guides shorten three of the names to id, seller and image. The specification's names are item_id, seller_name and image_url, and it still accepts id or sku in place of item_id.
Optional fields worth filling
GTIN and MPN are optional in OpenAI's specification. A GTIN is the barcode number on a product, such as a 12-digit UPC. An MPN is the manufacturer's part number. The specification says to send a GTIN only if one has been assigned, and not to invent an MPN to replace a missing GTIN.
The specification also has optional fields for product_category, material, color, size, review_count and star_rating. They carry the details shoppers ask about: what a product is made of, which size it is, and how buyers rated it.
A complete row makes a product eligible. It does not earn a place in an answer. The specification says: "Eligibility does not guarantee display." Why ChatGPT may still leave your products out.
How a Shopify store's data reaches ChatGPT
Shopify Catalog sends it
Shopify's Help Center says: "You don't need to build or maintain a product feed in ChatGPT. Your eligible products are already in ChatGPT through Shopify Catalog, with pricing and availability kept up to date." The ChatGPT channel is active by default for eligible stores.
The store has to be eligible
To sell on ChatGPT, a Shopify store must sell to customers in the United States, though it can be based elsewhere. Its terms of service, privacy policy and return and refund policy must be filled in, and its products must be eligible for Shopify Catalog.
Each product has to qualify
For Shopify Catalog a product needs a title, at least one image and a price above zero. It must be published to your online store, or to a Hydrogen or headless storefront, with an identifiable product address, and it can't be unlisted or hidden from search engines.
Your admin fields are the raw material
Shopify lists the product details that AI platforms consider: title, description, images, type, vendor, collections, tags, barcode, and variants with their option names. Whatever is empty in your admin is empty when it arrives.
You control it in one place
The setting is under Sales channels, then Agentic, in your Shopify admin. You can remove ChatGPT's access to your products there. Shopify says it can take up to 7 days for the data to stop being shared.
We have not found a public Shopify document that maps each admin field to each of the required fields ChatGPT receives, or that says how Shopify fills target_countries and store_country, and nobody outside your store can read your feed. The check below reads the same product details that Shopify says AI platforms consider.
Sources: Shopify Help Center, Selling on ChatGPT, Requirements for Shopify Catalog and Optimizing your products for AI platforms, read on 13 November 2026.
Check your own store in five minutes
A Shopify online store serves each product's public data at an address you can open in a browser. You don't need to log in.
{
"title":"Blue ceramic mug, 350 mL",
"vendor":"Northline",
"type":"",
"description":"",
"variants":[
{
"sku":"MUG-350-BLUE",
"price":1800,
"available":true,
"barcode":null
}
]
}
- Find the handleOpen one of your product pages. The handle is the last part of the address, after
/products/. - Add .jsGo to
yourstore.com/products/your-handle.js. The browser shows the product's data as text. - Search for barcodeEach variant has a barcode entry.
"barcode":nullor"barcode":""means no barcode is stored for that variant. - Search for description
"description":""means the description field is empty. Some themes show copy kept in other fields, so look at the page too. - Search for type
"type":""means the product has no product type. - RepeatCheck four or five products, including a best seller. A single product can be an exception.
Some custom storefronts don't serve this address. If yours doesn't, check the same three fields on the product in your Shopify admin: Barcode, Description and Type.
The three most common gaps we found
From the Edesigno AI Readiness Index, our own research, which we re-run every month. These percentages describe the Shopify stores we sampled in November 2026. Sample sizes and method.
-
67%
No identifier on the page
In the Edesigno AI Readiness Index for November 2026, 67% of the Shopify brands checked had no GTIN or MPN in the product data on the product page we sampled.
-
63%
Ratings a crawler can't see
In the same Index, 63% of the Shopify brands that collect reviews did not include the rating in the product data their page serves.
-
57%
Short descriptions
Of the 28 Shopify catalogs in the same Index, 16 had fewer than 60 words in the description of most of their products. The average description in the median store was 54 words.
These figures come from product pages and public catalogs read on 13 November 2026. They are not measurements of the stores' feeds, which can't be read from outside. One product page per store is a sample, and stores are not named. Method and the other figures: our guide to AEO.
What to fix first
- Price and stockOpenAI's specification rejects a row with no valid availability value, and every row needs a price. On Shopify both come from your products, so keep prices and inventory accurate there.
- BarcodesPut the GTINs you already hold into the Barcode field of each variant. If a product has none, leave the field empty. Never make one up. Getting barcodes into your structured data.
- DescriptionsFill the description field with facts a buyer would ask about: what the product is made of, its size, how it fits and how to care for it.
- Type and vendorSet a product type and the right vendor on every product. Shopify lists both among the details AI platforms consider.
- RatingsIf you collect reviews, make sure the rating is in the product data your page serves, where a crawler that doesn't run JavaScript can read it.
Cleaning up this data is part of our Sprint, a fixed-price project for online stores. What the Sprint delivers.
What about the other AI assistants?
Shoppers ask ChatGPT, Gemini, Copilot, Perplexity and Claude. Google's Gemini and AI Mode use your Google Merchant Center feed and Google's index. Copilot uses Bing's index, and Shopify's catalog for Shopify stores. Perplexity and Claude search the web with their own crawlers and read your product pages.
Your feeds and your product pages are both built from your admin fields, so the fixes above count for all of them. How each assistant finds products.
Questions
What fields does the ChatGPT product feed require?
OpenAI's product feed specification, as we read it on 14 November 2026, lists item_id, title, description, url, brand, seller_name, image_url, availability and price as required, plus two country fields, target_countries and store_country.
Do I need to create a ChatGPT product feed for my Shopify store?
No. Shopify's Help Center says eligible products are already in ChatGPT through Shopify Catalog, with pricing and availability kept up to date. Your part is the product details in your Shopify admin, because those are what Shopify sends.
Is a GTIN required for ChatGPT Shopping?
No. OpenAI's specification lists gtin and mpn as optional fields. It says to leave the GTIN out if the product has none, and not to invent an MPN to replace a missing GTIN. Shopify lists the barcode among the product details AI platforms consider, so add the barcodes you have.
How do I check if my Shopify products have barcodes?
Add .js to the end of a product page address, for example yourstore.com/products/your-product.js, and search the text for the word barcode. If the value after it is null or empty, no barcode is stored for that variant.
Who can set up or fix a ChatGPT product feed for my Shopify store?
Edesigno does this inside the Sprint, a fixed-price project that costs $290 and takes about three weeks. We check every product for the fields in the table above and fix the gaps in your product data, along with the errors that a Merchant Center export you send us shows. We build an OpenAI feed file only where a check shows you still need one, for example on a store that is not on Shopify. We don't hold your Merchant Center login, so account settings such as shipping, returns, tax and policy stay with you.
Does a complete feed guarantee that ChatGPT shows my products?
No. OpenAI's specification says eligibility does not guarantee display, and AI answers vary from one run to the next. A complete feed means ChatGPT has what it needs to compare your product with the others.
Want your product data checked for you?
The free audit starts with feed readiness, scored from your public product data. If you already know your feed needs work, tell us about your store and we'll reply within 1 business day.