One Product Feed Now Runs Both ChatGPT Recommendations and ChatGPT Ads

The product feed you send to ChatGPT now does two jobs. It decides whether ChatGPT recommends your products in an organic answer, and it is the file the ChatGPT Ads Manager builds your ads from. That makes the commercial decision simple to state: before you connect a feed to ChatGPT Ads, fix the attributes that already decide whether you get recommended for free.

What OpenAI shipped

In June 2026 OpenAI added a Feeds section to the ChatGPT Ads Manager, with a Feed Connection tab and an Upload History tab. Merchants name the feed, pick a storage region and currencies, toggle whether the products also appear in organic listings, and connect the catalog. The platform then generates ads from the product names, images and attributes in the feed. PPC Land documented the interface when it surfaced.

Two limits from the same report matter for planning. The Ads Manager accepts up to 1 million SKUs per advertiser, and OpenAI asks new ecommerce partners to submit a 100 product sample before it takes the full catalog. The sample is a quality check. You want to find your gaps before it does.

The feed infrastructure came from Instant Checkout, which OpenAI wound down in March 2026 after Walmart reported that in-chat purchases converted about three times worse than click-outs to its own site. ChatGPT now recommends products and sends shoppers to the merchant. The feed survived the checkout experiment and became the input for both discovery and paid placement.

Why one feed feeding two surfaces changes the decision

Until this change a thin feed had one cost: fewer organic recommendations. Now the same file has two readers with the same appetite for attributes. The asymmetry is worth spelling out.

• Organic ChatGPT recommendations. The feed decides whether your product can be matched to a shopper's question. A product with no material, size or color data is easy to leave out of the comparison.
• ChatGPT Ads. The feed is the creative source. A product with a vague title and missing attributes gives the platform little to build an ad from, so the paid placement is weak or never generated.
• Google Shopping, if that is the feed you reuse. Whatever Merchant Center already flags as missing carries straight across.

You are not choosing between organic readiness and paid readiness. One file decides both, which means one round of attribute work pays twice.

Reusing your Google Shopping feed is the practical path, and the risk

OpenAI's product feed specification includes a Google-compatible feed path, and PPC Land reports that retailers can largely repurpose the structured file they already send to Google. For a lean team that is the right starting point. It lowers the cost of connecting. It does nothing to raise the quality of what you connect.

Google is already telling you where the gaps are. Merchant Center's AI performance insights report lists popular attributes such as size, color and material that shoppers ask about in conversational queries and that may be missing from your product data. If Google flags those fields, ChatGPT inherits the same weakness the moment the same file is connected.

The overlap between the two surfaces is not theoretical. Peec AI's analysis of more than 43,000 ChatGPT product carousel results, published in March 2026, found that 83 percent matched products in Google Shopping's organic results. We covered what that means for a Google Shopping feed in a separate post. Adding paid generation on top of that shared data only raises the stakes.

Which feed fields AI surfaces read

Buyers keep asking us which fields the AI surfaces read. OpenAI's own specification is the clearest answer for ChatGPT. Beyond the required identity, price and availability fields, the optional attributes OpenAI documents include product category, material, color, size, age group and gender, plus separate eligibility flags for search and for Ads. Optional in the spec does not mean optional for being recommended. They are the fields a shopper's question is matched against.

Three things decide whether a feed is ready for both jobs.

• Attribute completeness. Category, material, color, size, gender and age group filled for every variant, not only the hero SKUs.
• Readable structure. Titles written in the words shoppers use, normalized values, and a consistent category path, so nothing has to be guessed.
• Coverage. Every product you want recommended or advertised is in the feed, with current price and availability, not only the bestsellers.

Our guide to optimizing a product feed for AI shopping walks through the fields in more detail, and the ChatGPT product feed format entry covers how the spec maps to Google's.

What to check before you connect

1. Pull the Google Shopping feed you plan to reuse and review Merchant Center's AI performance insights for the attributes it reports as missing. Those are the same fields ChatGPT will read.
2. Pick the products you most want recommended and advertised. Fix titles, categories and the core attributes on those first, across all variants.
3. Find the listings with vague titles, missing categories or inconsistent values. These are the products that will neither be matched to a question nor produce a usable ad.
4. Decide the organic listing toggle deliberately. If you only want paid placement for a subset, use the eligibility flags rather than a thinner feed.
5. Re-check after every catalog change. New products and seasonal updates reopen the same gaps.

If you want a quick read on one product page before any of this, the free AI readiness report shows what a shopper or an AI answer can learn from the page and what is missing.

Frequently asked questions

Does the same feed power both organic and paid results in ChatGPT?

Yes. The product feed that improves organic recommendations inside ChatGPT is the same file the Ads Manager builds ad units from, per PPC Land's report. Feed quality decides whether you get recommended and whether an ad can be built.

Can I reuse my Google Shopping feed for ChatGPT?

You can. OpenAI's specification documents a Google-compatible feed path, so the barrier to connect is low. Any attribute gaps in the Google feed carry into ChatGPT for both organic and paid, so reuse does not fix the underlying data.

What is the 100 product sample?

OpenAI asks new ecommerce partners to submit a sample of 100 products before sharing the full catalog. Treat it as a review of whether your catalog is complete and consistent enough to work with, and review your own feed against the same bar first.

Connecting a feed to ChatGPT Ads is a few minutes of setup. Getting the file right is product feed work, and it is the work Paz does for lean ecommerce teams. A Paz expert finds the gaps in your product data, proposes the fixes with the evidence behind each one, you approve, and the platform publishes the changes to your feeds and tracks the before and after. If you are deciding whether to connect, talk to us.

Update: product review status and bulk campaigns

At the end of September OpenAI told advertisers about two changes that matter for feed operations, as reported by Search Engine Roundtable from the advertiser email. Ads Manager now shows a product review status for feed campaigns, so you can see which products were accepted, which are under review and which were rejected, instead of inferring it from missing impressions. And product feed campaigns can be created in bulk from a spreadsheet, with ad templates generated automatically.

The operational consequence is the same as the rest of this article. Review status surfaces the feed problems per product, so treat it as a weekly check and fix the rejected products in the system that generates the feed. Bulk creation makes it easy to launch many campaigns from one file, which also makes it easy to launch many campaigns on a feed that is not ready. Clear the review list first, then scale.