OpenAI product feed
The OpenAI product feed is a structured product data specification that merchants submit so ChatGPT can index and display their products with current price and availability. It carries required fields including item_id, title, description, url, brand, image_url, price, availability, and seller details, plus eligibility flags controlling whether items can appear in search and in checkout. Its most consequential requirement is that product and variant identifiers stay stable over time.
What the feed is for
OpenAI's documentation frames it directly: provide a structured product feed so ChatGPT accurately indexes and displays your products with up-to-date price and availability. It is a grounding mechanism. Rather than depending on a crawl of your PDP, the merchant supplies the facts.
The feed powers product discovery in ChatGPT search results, checkout inside ChatGPT where enabled, ad eligibility where flagged, and accurate price and stock display. Delivery is by file upload or through API endpoints for feeds, products, and promotions.
The field set
Core required fields cover identity (item_id, title, description, url), product context (brand, image_url), commerce data (price, availability), merchant identity (seller_name, seller_url), and eligibility control (is_eligible_search, is_eligible_checkout). Some requirements are conditional: availability_date for pre-orders, seller policies when checkout is enabled.
The eligibility flags are more useful than they look. They let a merchant appear in discovery without opening checkout, which for a distributor with contract pricing and terms accounts is usually the only sensible configuration.
The identifier stability rule
The requirement that product and variant identifiers remain stable over time is the one to raise internally, because it is a governance commitment rather than a mapping exercise.
SKU identifiers get reassigned during ERP migrations, platform replatforms, warehouse re-racks, and catalog consolidations after an acquisition. Each of those events is normal, planned, and quietly breaks every downstream system that stored your identifier — including this feed, and including any agent that cached a reference to a product.
Freezing SKU IDs as immutable, and treating renumbering as a data migration with mappings rather than an overwrite, is the practical answer. It costs almost nothing to decide before a migration and is expensive to reconstruct after one.
How it compares to a Google Shopping feed
Similar shape, different specialization. A Google Shopping feed is mature, deeply specified for retail merchandising, and carries taxonomy and unit-pricing attributes that industrial catalogs can lean on: product_detail accepts section_name : attribute_name : attribute_value triples for technical specs, and unit_pricing_measure, multipack and is_bundle keep pack concepts distinct.
The OpenAI feed is newer and leaner, and its checkout flags reflect its integration with agentic checkout.
For most distributors neither feed is the first investment. Both are export formats over the same underlying record, and a feed built on a catalog with forty percent attribute coverage exports forty percent attribute coverage. Fix the record; the feeds are then a mapping job.
Frequently asked questions
What is the OpenAI product feed?
A structured product data specification merchants submit so ChatGPT can index and display their products with current price and availability. It supports product discovery in ChatGPT search, checkout where enabled, and ad eligibility, and is delivered by file upload or through API endpoints.
What fields does the OpenAI product feed require?
Core required fields include item_id, title, description, url, brand, image_url, price, availability, seller_name and seller_url, plus the eligibility flags is_eligible_search and is_eligible_checkout. Some fields are conditionally required, such as availability_date for pre-orders and seller policies when checkout is enabled.
Why does identifier stability matter so much?
Because the specification requires product and variant identifiers to remain stable over time, and normal operational events break that. ERP migrations, replatforms and post-acquisition catalog merges routinely reassign SKU IDs, which invalidates every stored reference including cached agent references.
Is the OpenAI feed different from a Google Shopping feed?
Yes, though they overlap heavily. The Google specification is more mature and richer in retail merchandising and unit-pricing attributes, while the OpenAI feed is leaner and includes checkout eligibility flags tied to agentic purchasing. Both are exports of the same underlying product record.