Skip to guide content

ChatGPT product feed requirements

ChatGPT product feed requirements, made practical.

Know what to include, how to map your catalog and what to check before submitting. Use the field reference and preparation checklist to turn a feed project into clear next steps.

See what your catalog is missing.

No login to start · No card · No advertiser API key

Public catalog review. See accessible product-data issues and examples before creating an account.

01 Make product details useful02 Keep every offer consistent03 Review changes with confidence
The practical answer

A ChatGPT discovery feed needs product identity, factual product content, public product and image URLs, current offer data and market information. Variants have additional requirements. Always check OpenAI's current specification and your integration's supported markets.

By the Magnify teamReviewed Practical merchant reference
01

Which fields does a ChatGPT product feed need?

Use the field checklist below to review your source data. Keep IDs stable, use real brand and seller values, and keep product links and images publicly accessible. Send prices with currency and an explicit stock status supported by the destination.

Required fields need actual values. Do not fill gaps with invented brands, guessed specifications or placeholder text. An unknown optional fact is a source-data task; fabricating it makes the feed less reliable. Refer to the official specification for exact formats and conditional requirements.

From your catalog to the feed

The field checklist, without the guesswork.

Review the source value and the actual offer—not just the column name. These are discovery requirements, not a complete ads or checkout contract.

Required discovery fields · reviewed 18 November 2026
OpenAI fieldPossible sourceWhat to check
item_ididKeep the same identity for the same item or variant.
titletitleName the actual product and selected variant.
descriptiondescriptionUse factual product details supported by your store.
urllinkOpen the correct product and variant on a public page.
brandbrandMatch the real brand shown on the product page.
seller_nameMerchant recordIdentify the actual seller supplying the offer.
image_urlimage_linkUse a public image showing the correct variant.
availabilityavailabilityMap the live stock state to an accepted value.
pricepriceInclude currency and match the live offer.
target_countriesMarket configurationConfirm the countries allowed for this integration.
store_countryMerchant configurationConfirm store and market setup with the destination.

Mappings are starting points. Variants have additional requirements. Confirm exact formats and supported markets in the current official specification.

02

Give every purchasable variant a stable identity

A size or color selection can change the price, stock or image. Represent the actual selection in its own item record. Preserve that item identity when the offer changes, rather than creating a new ID every time you edit a title.

For a variant group, review group_id, listing_has_variations and variant_dict in the current reference. Use a shared group identity, distinct item identities and consistent option names. Check that each row points to the right selection and carries that selection's offer details.

Want to see this in your own catalog?

Start with your store URL. No login to begin.

Get my free catalog review →
03

How to map a Google Merchant Center feed

Reuse your verified catalog values, but review the destination format explicitly. A product link and an image link may use different field names; a value can also have different accepted formats. Brand, seller, variant and market information may need a separate mapping.

Check more than the column headers. Open a sample product URL, inspect the selected variant and compare the live offer with the row. Keep a record of each mapping and its source so future changes can be reviewed without starting over.

04

How to submit a product feed to ChatGPT

Use the merchant workflow available to your business. OpenAI's current submission guide starts in Business Hub: claim the business, complete the identity steps shown for your account and open the products section. Prepare the catalog using the current specification.

Choose a supported delivery route: manual upload, a hosted HTTPS URL or SFTP, as available in the portal. Review the upload diagnostics, correct errors and send refreshed snapshots as offers change. A catalog review is useful preparation, but it does not perform these submission steps for you.

05

Check market support and keep changing offers current

Do not interpret a country column as worldwide access. The current reference says standard OpenAI-format uploads target the US; additional countries and currencies need confirmation for the integration. Verify the market setup available to your business before preparing regional files.

Price and inventory can change after an upload. Keep the source catalog and the delivered snapshot current, and verify the live offer when a shopper reaches your store. A feed is not a reservation, a cart update or a purchase confirmation.

06

Discovery, paid ads and checkout have separate requirements

A discovery feed does not automatically enable advertising or checkout. Those surfaces have their own eligibility and integration requirements. Do not treat an advertiser API key or an accepted product upload as proof that every other capability is ready.

Use this checklist to prepare product data for shopping discovery. Use the separate ChatGPT Ads hub when you need paid advertising. For checkout, follow the integration instructions applicable to your business.

Take this to your next catalog review

Your ChatGPT shopping feed preparation checklist.

Tick the items you have checked. This working list is not a validation score and does not save or upload your answers.

Download the checklist

Free text file · No email required · Reviewed 18 November 2026

Before you get started

Frequently asked questions

Which product feed fields are required?

The current discovery reference lists item_id, title, description, url, brand, seller_name, image_url, availability and price, plus target_countries and store_country. Variants and other integrations add conditional requirements. Check the current official reference before submission.

Can I submit a Google Shopping feed unchanged?

Do not assume so. Reuse the source data, then check accepted field names, values, identities, variant grouping and markets against the destination specification. Test the actual output and review upload diagnostics.

Does the standard feed support every country?

No. The current specification says standard OpenAI-format uploads target the US. Confirm support for additional markets and currencies with OpenAI for your integration.

How often should I refresh the feed?

Refresh it to reflect changes in your source catalog, especially price, stock, URLs and variants. Use the cadence and delivery behavior supported by your merchant setup. Do not rely on stale snapshots for changing offers.

Does the free Magnify review validate the full OpenAI schema?

No. It reviews publicly accessible catalog data and identifies common quality issues. Complete feed validation and OpenAI's upload diagnostics remain separate checks.

Official references

Reviewed 18 November 2026. Requirements can change; the current OpenAI documentation is the authority for your integration.

Product facts, made useful

A better feed makes the difference clear.

Your customers see a product. Your catalog needs to explain it. Start with factual titles and attributes that help someone compare.

Red running shoe
Footwear
Basic catalog title

Runner 2.0

Enriched feed example

Lightweight Running Shoes — Red — Mesh Upper

Product type · color · material

Green sofa in a living room
Furniture
Basic catalog title

Sofa 103

Enriched feed example

Modern 3-Seater Sofa — Forest Green

Product type · seating capacity · color

Table lamp in a home interior
Home decor
Basic catalog title

Lamp 22

Enriched feed example

Bedside Table Lamp — Warm White Light

Product type · use case · lighting

Illustrative examples. Verify every attribute against your actual products. Improved data does not guarantee a recommendation or sale.

Know what to work on next

From “is my feed good enough?” to a useful first review.

Start with a store URL. Magnify reviews publicly accessible product data so you can see issues and suggested examples before signing up.

  • See where product information is thin or unclear.
  • Review suggested examples alongside the source data.
  • Decide which catalog changes deserve attention first.
Review my catalog — free

Available checks depend on accessible data. This is not complete OpenAI feed validation, submission or approval.

Your catalog is the starting point

Find the gaps.
Make the next edit count.

See what your public product data can tell you before building another feed or buying another tool.

Review my catalog — free

No login to start. No automatic changes to your catalog.