Getting your products discovered in ChatGPT is no longer only a question of whether your website can be crawled.
E-commerce brands can provide structured product data directly to OpenAI through ChatGPT product feeds, giving the platform a clearer and more current picture of what you sell, how much it costs, whether it is available, and which version of a product a shopper should actually see.
A product page gives people the full buying experience. A product feed gives a system the information it needs in a more structured format.
As of September 2026, OpenAI’s stable discovery specification requires nine core fields, with optional attributes available for variants, shipping, returns, reviews, product identifiers and other commerce information.
This guide covers what a ChatGPT product feed does, what information you need to include, and the most important things to get right before submitting one.
What Is a ChatGPT Product Feed?
A ChatGPT product feed is a structured dataset containing information about the products an e-commerce business sells so OpenAI can use that information for product discovery within ChatGPT.
Instead of relying entirely on crawling individual product pages, a feed can provide structured details such as product ID, title, description, URL, brand, image, availability and price.
It does not guarantee that a product will be recommended, but it gives OpenAI clearer and more current information about the catalog.
If you want to understand the broader recommendation process first, read The Darl’s guide to how ChatGPT recommends products.
Do You Need a Product Feed If ChatGPT Can Already Crawl Your Website?
Not necessarily.
OpenAI says merchants do not need a product feed simply because ChatGPT can crawl their websites. A feed does, however, give merchants more control over the product information ChatGPT can use and can help keep details such as price, availability and variants current.
That becomes especially useful for large catalogs.
A retailer with thousands of products may have changing stock, different variants, promotional content, review widgets and several URLs connected to the same item. A structured feed reduces much of that ambiguity.
Instead of asking a system to work out which image belongs to the blue size-10 version of a shoe, you provide that information directly.
How Do You Submit Products to ChatGPT?
OpenAI’s merchant workflow allows businesses to make structured product data available through its commerce infrastructure.
For most stores, the preparation process looks like this:
- Audit your existing product data.
- Map the catalog to OpenAI’s product-feed specification.
- Create a record for each purchasable product or variant.
- Validate IDs and required fields.
- Check that product and image URLs work.
- Make sure price and availability match the live store.
- Connect or submit the feed through the available merchant workflow.
- Keep the feed updated.
OpenAI has also stated that Shopify and Etsy catalogs are already connected to its shopping ecosystem, so merchants on those platforms may not need to build a separate connection from scratch.
Which Fields Are Required in a ChatGPT Product Feed?
OpenAI’s current stable discovery specification requires nine fields.
| Field | What It Represents |
| item_id | Stable unique ID for the product or variant |
| title | Product name, including variant where relevant |
| description | Factual product description |
| url | Product detail page URL |
| brand | Product brand |
| seller_name | Seller supplying the offer |
| image_url | Main product image |
| availability | Current stock status |
| price | Regular product price and currency |
The difficult part is not creating nine columns. It is making sure those fields describe the same product consistently across the feed, product page and other commerce systems.
Item ID
The item_id is the stable identifier for an individual item or variant.
A stable ID allows systems to keep track of the same product even as details such as price, stock status or title change over time.
If your store already has reliable SKUs or product identifiers, use those rather than inventing a separate ID system only for AI.
Product Title
The title should clearly identify the item and, where relevant, the selected variant.
A title such as “Hydrating Mineral SPF 40 — Tinted, Medium” gives the system much more useful information than “Our Bestselling Sunscreen,” because it identifies the actual product and variation rather than relying on context that may not exist in the feed.
Product Description
Descriptions should be factual and easy to understand.
For a skincare product, useful information might include the product type, intended use, relevant ingredients, size, formulation and finish.
The goal is not to stuff keywords into the feed. It is to describe the product clearly enough that the system can understand what it is.
This principle also applies beyond product feeds. The Darl’s guide to Generative Engine Optimization explains why consistent entities and clearly structured information matter across AI discovery systems.
Product URL
Each item needs a public product detail page URL.
Where possible, variant records should link directly to the relevant selected option. If the feed describes a black size-10 shoe but the destination page opens on a blue size-8 version, the experience becomes confusing.
Brand and Seller Name
Brand and seller are separate fields because they are not always the same entity.
The brand should match the name customers see on the product page. If “Luma Skin” is the public-facing brand, use that consistently instead of switching between the brand name and a parent company or billing entity.
Keeping the naming consistent makes it easier for systems to associate products with the right brand.
Image URL
The image should represent the actual product or variant in the record.
If the record describes a black handbag, the main image should not show the beige version.
For visual categories such as beauty, apparel and home goods, this consistency matters even more.
Availability
OpenAI supports availability values such as:
- in_stock
- out_of_stock
- pre_order
- backorder
- unknown
The feed should reflect the real stock status whenever possible. A feed that continues to report an item as available after it has sold out quickly becomes less useful.
Price
Price should include the item’s regular price and currency and should match the buying destination.
If your website changes price frequently, the feed needs an update process rather than a one-time upload.
How Should Product Variants Be Structured?
OpenAI recommends creating a separate record for each purchasable variant.
Fields such as group_id and variant_dict can connect related variants while preserving their individual product information.
For example, a shoe sold in black and white across several sizes should not be represented as several unrelated products. Each purchasable option can have its own item_id, while related variants remain connected to the same product family.
This makes it possible to keep price, availability, URLs and images accurate for each option.
What Optional Product Data Should You Add?
The minimum required fields are enough to create a valid discovery feed, but additional attributes can make products easier to understand.
OpenAI supports information such as:
- GTIN and MPN
- product category
- material
- color
- size
- dimensions
- weight
- additional images
- shipping
- returns
- reviews
You do not need to fill every optional field simply because it exists.
Use attributes that help people distinguish the product when making a buying decision. Shade matters for lipstick, dimensions matter for furniture, and size and color are essential for footwear.
Can You Use an Existing Google Product Feed?
In many cases, yes.
OpenAI’s current documentation supports Google-compatible product feeds, which means brands already maintaining good Google Merchant Center data may have much of the necessary product information in place.
That does not mean you should submit the existing feed without checking it.
Make sure IDs are stable, variants map correctly, product titles are accurate, URLs work, prices match the site, stock is current and brand names remain consistent.
Existing feed infrastructure can save work, but it still needs QA.
What Are the Most Common ChatGPT Product Feed Mistakes?
Most problems are not mysterious AI issues. They are standard commerce-data problems.
Common mistakes include:
- changing IDs when product details change
- combining several purchasable variants into one record
- using an image that does not match the variant
- letting the website and feed show different prices or stock status
- writing overly promotional or keyword-heavy descriptions
- assuming that submitting a feed guarantees a recommendation
A clean feed makes your products easier to understand. It does not guarantee they will be displayed.
Will a Product Feed Make ChatGPT Recommend My Products?
Not automatically.
Submitting a product feed can improve the accuracy and completeness of the product information OpenAI has available, but recommendations depend on more than the presence of a feed.
Product relevance, user intent, product information and available alternatives can all influence what gets shown.
A feed answers one question:
Does the system have accurate structured information about what we sell?
The wider AI visibility question is different:
Does the broader information ecosystem give AI systems enough context to understand and potentially recommend the brand?
For that second layer, see The Darl’s guide to AI SEO services.
Product Feed vs Product Schema
They are related, but they are not the same thing.
| Data Layer | Main Role |
| Product page | Human-readable buying destination |
| Product schema | Machine-readable information attached to the page |
| Merchant/product feed | Structured catalog sent to commerce platforms |
| ChatGPT product feed | Structured product data used for ChatGPT commerce discovery |
These layers should agree with each other.
If one source shows a different price, product name or stock status, the problem is not simply technical SEO. It is inconsistent product data.
Can Customers Buy Directly Inside ChatGPT?
OpenAI’s commerce strategy has changed over time.
As of September 2026, ChatGPT shopping is focused heavily on product discovery, while purchases are generally completed through merchant websites or apps unless a deeper integration is involved.
For marketers, this makes the destination page just as important as the AI discovery layer.
ChatGPT may introduce the product, but the merchant still needs to close the sale with accurate pricing, good imagery, clear information, credible reviews and an easy checkout experience.
A Practical Product Feed QA Checklist
Before a feed goes live, check four things.
Product identity
- Does every product or variant have a stable ID?
- Are variants grouped correctly?
- Does the brand name match the site?
Product information
- Does the title clearly identify the item?
- Is the description factual?
- Does the main image match the record?
Commerce accuracy
- Does price match the product page?
- Is availability current?
- Does each URL open the correct product or variant?
AI discovery readiness
- Are the important product attributes present?
- Does the feed agree with the website and structured data?
- Is the same product represented consistently across other channels?
What Should Shopify Brands Do?
Shopify merchants should start by cleaning the catalog they already have.
Review titles, descriptions, variants, IDs, prices, availability, images, categories and canonical product URLs.
The goal is not to create a second catalog for AI. It is to make the existing catalog reliable enough that the same product information can travel across multiple systems without changing meaning.
What Should Larger E-commerce Teams Do?
Larger brands should avoid building a separate manually maintained “ChatGPT feed” if they already use a PIM, ERP or central commerce platform.
Instead, identify the system that already holds the most reliable product, pricing, inventory and media data, then map that source to the ChatGPT specification.
The long-term goal should be synchronization rather than another isolated catalog.
Frequently Asked Questions
How do I submit products to ChatGPT?
Merchants can make structured product information available through OpenAI’s commerce infrastructure and supported merchant workflows. The first step is making sure the catalog complies with the current product specification.
Does ChatGPT crawl my product pages?
Yes, ChatGPT can discover product information from the web. A product feed is not always required, but it gives merchants more direct control over details such as price, availability and variants.
Which fields are required in a ChatGPT product feed?
The current discovery specification requires item_id, title, description, url, brand, seller_name, image_url, availability and price.
Does submitting a feed guarantee that my products appear in ChatGPT?
No. A feed makes structured product information available, but it does not guarantee display or recommendation.
Can I use my Google Merchant product data?
OpenAI supports Google-compatible feed formats, so existing Merchant Center data can be a useful starting point. It should still be checked against the current ChatGPT product-feed specification before submission.
The Bottom Line
A ChatGPT product feed is not just another catalog export.
It is a structured way to help an AI system understand what you sell, which version is available, how much it costs and where a shopper should go next.
The brands best prepared for AI-driven product discovery will not simply have the most content. They will have reliable product data, consistent product entities and a commerce experience that carries the same information from discovery through purchase.
If your product feed is clean but the wider brand is difficult to understand, there is still work to do.
The Darl connects SEO, AI visibility, content and commerce strategy so those systems do not operate as separate projects. Explore our full marketing services or learn more about our AI SEO services.
For e-commerce brands selling through marketplaces as well, The Darl’s Amazon Services can help connect marketplace visibility with the wider search and brand strategy.
