ChatGPT's Ad Manager Just Made One Feed Do Two Jobs

OpenAI moved product feed management into a dedicated ChatGPT Ads Manager, first surfaced June 2, 2026. The same catalog file that decides whether ChatGPT recommends your products organically now also generates your paid ads. One feed, two doors. If that file is thin, you lose on both.

TL;DR: OpenAI's new ChatGPT Ads Manager auto-generates ads straight from your product feed, up to 1 million SKUs per advertiser. That feed already controlled organic discovery inside ChatGPT. Now it controls paid placement too. A weak feed makes you invisible on the free side and ungenerable on the paid side. The fix is feed quality, measured before you connect.

What did OpenAI actually ship?

OpenAI added a dedicated Feeds section to a ChatGPT Ads Manager interface, with Feed Connection and Upload History tabs. Merchants connect a catalog, set eligibility filters, and the platform auto-generates ads from product names, images, and attributes in the feed. The structured data file is the creative source. There is no product-by-product campaign building.

The detail was first surfaced by entrepreneur Juozas Kaziukenas on LinkedIn and built on earlier reporting from Digiday and PPC Land. The system handles up to 1 million SKUs per advertiser. New e-commerce partners submit a 100-product sample before sharing the full catalog. At least one brand onboarded through Criteo, OpenAI's first ad-tech partner, and called it more straightforward than expected.

The piece most retailers will miss: this same feed was already the input for organic product discovery inside ChatGPT. OpenAI did not build a separate ad catalog. It pointed the ad engine at the catalog you already maintain.

Why does one feed feeding two surfaces change the math?

Until now, a brand could rationalize a thin or incomplete feed. Instant Checkout had been discontinued in March 2026 after Walmart disclosed that in-chat conversion ran about three times lower than click-out to its own site, part of the same shutdown PPC Land documents. With the buy button gone, feed gaps felt survivable. That excuse is now expired.

The product feed is the single most consequential asset inside ChatGPT. It controls organic surfacing and paid ad eligibility at the same time.

Here is the asymmetry that matters. On the organic side, a thin feed means ChatGPT has nothing to rank you on, so you do not get recommended. On the paid side, the platform builds the ad from your feed's names, images, and attributes, so a thin feed cannot generate a usable ad at all. The same missing color, material, and size attributes hurt you in both places. You are not choosing between organic and paid readiness. The feed decides both, from one file.

This is the structured-data convergence that has been building all year. The way you structure product data for AI agents is no longer a back-office task. It is the lever for visibility across every surface ChatGPT touches.

Doesn't reusing my Google Shopping feed solve it?

OpenAI accepting the Google Shopping feed format lowers the barrier to entry. It does not raise the ceiling on quality. Whatever gaps live in your Google feed today carry straight into ChatGPT, on both organic and paid.

Google itself now flags those gaps. Its Merchant Center AI insights highlight missing color, material, and style attributes as the fields most likely to suppress performance. If Google is telling you those attributes are weak, ChatGPT inherits the same weakness the moment you connect the same file. The ChatGPT product feed format borrows heavily from Google's spec, which is convenient and also a trap: convenience hides the fact that a mediocre feed produces mediocre results twice.

Surface	What the feed controls	What a thin feed costs you
Organic ChatGPT discovery	Whether your product gets recommended	No recommendation, no traffic
ChatGPT Ads Manager (paid)	Whether an ad can be auto-generated	No ad creative, no paid placement
Google Shopping (source)	Feed quality flagged by Merchant Center	Suppressed performance, inherited by ChatGPT

A study of 43,000-plus listings found that 83% of ChatGPT product carousel results matched Google Shopping organic results, which is why your feed already powers both platforms. Adding paid generation on top of that shared feed only raises the stakes on the underlying data.

How do you know if your feed is ready before you connect it?

You measure it. The 100-product sample OpenAI requires from new partners is a quiet quality gate: it is OpenAI checking whether your catalog is coherent enough to work with before it grants full access. You want to find your gaps before that sample does.

Three things decide feed readiness for both organic and paid:

Attribute completeness. Color, material, size, style, gender, and category fields filled for every variant, not just the hero SKUs.
Machine-readable structure. Clean titles, normalized values, and consistent taxonomy so the platform can parse and rank without guessing.
Coverage depth. Every SKU you want surfaced is in the feed, not just your bestsellers. The AI product found rate measures exactly this gap between what you sell and what AI can see.

Product feed optimization for AI is the discipline here. It is not a one-time export. It is ongoing measurement of how complete and parseable your catalog is across the surfaces that now read it, which is the full scope of ChatGPT shopping discovery and, now, paid generation.

What to Do This Week

Pull your current Google Shopping feed and audit the top three attribute gaps. Check color, material, and style coverage across all variants, not just bestsellers. These are the fields Google flags and the ones ChatGPT inherits.
Count your machine-readability failures. Find SKUs with vague titles, missing categories, or inconsistent taxonomy. These are the listings that will neither rank organically nor generate a usable ad.
Score your feed before the 100-product sample does. Run a readiness check so you know what OpenAI's gate will catch. Tools like our AI Readiness Report measure feed completeness against AI surface requirements.
Prioritize the SKUs you most want surfaced. Fix attribute gaps on those first, since the same fields drive organic recommendation and paid ad creative.
Re-audit after every catalog change. A feed is not static. New products and seasonal updates reopen the same gaps you just closed.

Frequently Asked Questions

What is the ChatGPT Ads Manager?

It is OpenAI's interface for running paid ads inside ChatGPT. As of June 2026 it includes a Feeds section that connects a merchant's product catalog and auto-generates ads from product names, images, and attributes, handling up to 1 million SKUs per advertiser.

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

Yes. The product feed that improves organic product discovery inside ChatGPT is now the same file that feeds the Ads Manager's auto-generated ad creative. Feed quality determines both whether you get recommended and whether an ad can be built.

Can I reuse my Google Shopping feed for ChatGPT?

You can. OpenAI accepts the Google Shopping feed format, so the barrier to connect is low. The catch is that any attribute gaps in your Google feed carry into ChatGPT for both organic and paid, so reuse does not fix underlying quality problems.

What is the 100-product sample requirement?

OpenAI requires new e-commerce partners to submit a 100-product sample before sharing their full catalog. It functions as a quality gate, letting OpenAI confirm a catalog is coherent and parseable before granting access to feed up to 1 million SKUs.

Did OpenAI bring back in-chat checkout with this?

No. Instant Checkout was launched in September 2025 and discontinued in March 2026. ChatGPT now recommends products and directs users to the merchant's site to complete the purchase. The Ads Manager is about discovery and paid placement, not checkout.

Why does attribute completeness matter so much now?

Because the same missing attributes hurt you twice. A feed without complete color, material, and size data cannot rank well organically and cannot generate a strong ad. The penalty for a thin feed used to be lost discovery; now it is lost discovery and lost paid eligibility.

The buy button is gone from ChatGPT, but the feed never mattered more. OpenAI just turned your catalog into the single input for both how you get found and whether you can pay to be found. The brands that treat the feed as a living, measured asset will show up on both doors. The ones still exporting a stale file once a quarter will quietly disappear from both.