ChatGPT Product Data: What Sellers Need to Submit
Sellers can use built-in platform catalogs, crawlable public pages, or an approved direct feed to give ChatGPT accurate product data, and the two-path check reveals which route their store supports.
ChatGPT gets product data from merchant or commerce-platform feeds, but it can also read public store pages. Shopify and Etsy catalogs are already built in. Sellers on other platforms can qualify through public pages and apply for direct-feed access.
I've checked the current OpenAI guidance for my two Shopify stores, and its merchant page says Shopify already supplies their catalog data. I still need to confirm OAI-SearchBot can load a live product URL and that its price, stock, variant, and ID match the catalog record. The checks below show which route fits your store and what to fix first.
Key takeaways
What is ChatGPT product data, and how does it get there?
ChatGPT product data is catalog details from merchant feeds, commerce platforms, and public store pages. OpenAI says its shopping tool can use Agentic Commerce Protocol merchant data and public product information. Both sources may describe the same item. They still reach ChatGPT through separate systems and refresh at different times.
A feed uses a set catalog from a platform, provider, or onboarded merchant. That lets OpenAI check each record and use current prices and stock. Public-page search reads store pages instead. It can cite or surface a product without a direct feed, but the page must be readable and relevant to the request.
Your platform sets the first task because OpenAI's shopping guidance says Shopify product data already reaches ChatGPT through Shopify Catalog.
OpenAI's merchant page says Etsy catalogs also need no extra setup. Sellers elsewhere can use public pages as they apply to share a feed.
Catalog access only puts an item within reach, so verify the AI product suggestion before spending on it. I would start with the route OpenAI already has. Rebuilding it adds another place for prices, stock, or IDs to drift. A second feed makes sense when your current platform omits products or can't meet the update schedule agreed during onboarding.
Run the Two-Path Readiness Check on your own store
These two checks test public-page access and feed coverage as separate routes into ChatGPT. A store may qualify through either route.
Run these two checks,
- Public-page path: Confirm OAI-SearchBot can reach and read one live product page.
- Feed-backed path: Confirm platform coverage or direct-feed onboarding and required fields.
Start with the public page, then confirm how the catalog reaches OpenAI.
1. Check the public-page path
Your product page passes when OAI-SearchBot can reach it and shoppers can read the key facts. OpenAI's search guide requires crawler access, and your host or content delivery network must also allow its published IP addresses.
Check one live product URL in this order,
- Open /robots.txt and look for a block that names OAI-SearchBot.
- Remove any Disallow rule that covers the page you want eligible.
- Check the firewall, bot service, and image host for blocked OpenAI traffic.
An open robots file can't override a server error.
Once access is clear, review the page in a private browser window while a successful server response confirms crawler access. The visible details should include the product name, selected variant, price, stock, main image, seller, shipping terms, and return terms. Keep them visible without an account or location prompt.
Product schema can repeat those facts for machines. OpenAI's public search guide makes crawler access the eligibility requirement. Use schema to confirm that the structured facts match the page.
I use the live page as the source of truth for shopper-visible facts. ChatGPT may fetch that page at a different time from a feed refresh. When the two show different prices, find the system publishing the stale value and correct that record.
2. Check the feed-backed path
Shopify and Etsy sellers already have feed coverage, while other sellers need OpenAI onboarding before sending a direct feed. This coverage doesn't require an application. Results still depend on accurate product names, variants, images, prices, and stock.
If you aren't on either platform, check whether your provider already has an OpenAI integration before building an export. Apply directly when your current route omits products or can't refresh price and stock often enough. OpenAI confirms provider and file support during onboarding. Applicants now join a waitlist, and OpenAI plans a self-service merchant portal later in 2026.
OpenAI's discovery file requires these fields,
- item_id
- title
- description
- url
- brand
- seller_name
- image_url
- availability
- price
Each row in OpenAI's product reference is one item or variant. Send full snapshots over SFTP at least daily in the approved format, using Parquet when available or compressed JSONL, CSV, or TSV. These fields support discovery, while checkout and ads use separate integrations.
Keep IDs stable when price or stock changes. Apply the same rule to channel-required product IDs. Updates will then attach to the correct variant instead of creating duplicate records.
Other channels need more fields, which our cross-platform feed checklist covers. For ChatGPT, use OpenAI's reference and your onboarding guide.
I build against the supported file-upload reference and the onboarding guide. The wider Agentic Commerce Protocol feed spec covers more of the protocol and uses other names and formats. A hybrid file may fit neither path.
When feed access or product data conflicts
Use OpenAI's supported file-upload reference for discovery feeds. Treat public-page access as eligibility. Handle checkout as a separate integration. The developer reference defines supported file uploads. The broader ACP page covers other commerce functions, including checkout fields. It uses different names and formats. Follow the source and onboarding instructions for your route.
OpenAI's merchant page says direct-feed applicants are on a waitlist while it builds a self-service tool. It also says purchases now default to merchant-owned sites or apps. OpenAI is moving away from a standalone Instant Checkout experience. A discovery feed can improve catalog coverage and freshness, but it doesn't switch on checkout by itself.
A manual ChatGPT query is a useful spot check when it cites the intended merchant page with the correct variant, price, and availability. Its scope is one response, so full catalog coverage and future placement remain unconfirmed. Fix catalog and page disagreements first, and apply for direct access when the current route omits products or misses your required update schedule.
FAQ
Do I need a product feed to show up in ChatGPT at all?
ChatGPT can use a public retail page even when the merchant doesn't have a direct feed. A feed can provide more complete and current catalog data.
What happens if my store isn't accepted for feed submission?
Keep OAI-SearchBot unblocked and maintain accurate public product pages while you wait. Public-page eligibility remains available when direct-feed access is delayed or declined.
Can I test whether ChatGPT already knows about my products?
Ask ChatGPT for the exact product and inspect the cited merchant page, then treat the result as a spot check. A useful match has the intended merchant, variant, price, and availability. Results may differ for another query or shopper.
