Amazon Ads is selling into ChatGPT, and your EU Sponsored Products budget should not move.

Amazon Ads is now selling into ChatGPT, and the correct action for an EU Sponsored Products seller today is none. On 10 September 2026 Amazon Ads announced a partnership with OpenAI to support advertising in ChatGPT, enabling advertisers to extend Amazon Ads campaigns into a conversational ad experience. Amazon's own announcement says select advertisers in the U.S. are currently testing ChatGPT Ads as part of a pilot. That is the whole of the confirmed fact, and it is worth reading twice, because the gap between what was announced and what most of the coverage this week will imply is where sellers lose money.

What Amazon actually confirmed

Three specifics, because they bound the story tightly. First, this is a partnership to support advertising in ChatGPT — an extension of existing Amazon Ads campaigns into a conversational surface, not a new campaign type inside your Sponsored Products console. Second, the announcement is filed against Amazon DSP; the quote in it is from Chris Conetta, Director of Omnichannel Supply at Amazon DSP. Third, participation today is a pilot, and the pilot is with select advertisers in the U.S.

Nothing in the announcement says Sponsored Products. Nothing in it says Germany, France, Italy or Spain. Nothing in it gives a general-availability date, a rate card, or a reporting spec. If you sell on Amazon.de and run Sponsored Products, the number of levers this gives you this morning is zero — and a vendor telling you otherwise this week is selling you a roadmap, not a feature.

The surface you would be buying, on OpenAI's own terms

OpenAI has published how ChatGPT ads are bought, and it is the more useful half of the picture. The pilot began on a CPM basis; OpenAI has since added cost-per-click bidding, launched a beta self-serve Ads Manager, and shipped Conversions API and pixel-based measurement. Advertisers can also buy through partners — agency partners including Dentsu, Omnicom, Publicis and WPP, and technology partners including Adobe, Criteo, Kargo, Pacvue and StackAdapt.

One line in that documentation matters more than the rest, and it is the one nobody will quote back at you in a sales call:

Partners help support campaign budgeting, bidding and advertising creative, while OpenAI's ads system controls all delivery decisions. Advertisers receive aggregated performance insights, without access to individual conversations.

Read that as an operator rather than a marketer. You set a budget and a bid. Someone else's system decides delivery. You get performance back in aggregate. That is a perfectly reasonable design for a privacy-preserving conversational surface — and it is close to the structural opposite of how a profit-first Amazon bidding loop works, where the whole exercise is attributing an outcome to a specific query, at a specific placement, on a specific ASIN, and moving one bid because of it.

Why this is a DSP story, not a Sponsored Products story

The distinction is not pedantry, it is where your budget lives. Sponsored Products is an auction against shopper intent already expressed inside Amazon, with keyword- and ASIN-level reporting you can bid against. DSP is programmatic reach across Amazon's properties and the open internet, bought against audiences. Conversational ad inventory in ChatGPT is, structurally, more supply for the second thing.

Which means the honest framing of this announcement for a mid-size EU seller is: Amazon added a new place to spend DSP money, in a market you are probably not in, at a stage that has no published benchmarks. If your Sponsored Products account is not already clearing its break-even ceiling on every marketplace you run, adding an untested top-of-funnel surface is not a growth strategy — it is a distraction with a launch date. The prior question is the one in what a good ACoS actually is in 2026, and it has not changed.

The part that is genuinely worth watching

There is a real second-order signal here, and it is not the ad units. It is that Amazon is now buying shopper attention at the point where the shopper is comparing options in language, before they reach a search box. We wrote about the on-Amazon version of that shift in what Rufus and Alexa mean for PPC and in agentic ads. This is the same movement continuing off-platform.

If discovery keeps migrating into conversation, the long-run consequence for Sponsored Products is not that it disappears. It is that the queries reaching Amazon's search box get progressively more decided — shorter funnels, higher intent, and comparison work already done elsewhere. That would move conversion rates and clearing prices on your existing keywords well before it ever gives you a new campaign type to buy. Which is the opposite of the advice you will hear this week: the change worth preparing for shows up in your current reports, not in a new line item.

What an EU seller should do about it

1. Change nothing in your live campaigns because of this announcement. There is no EU availability and no Sponsored Products surface to change.
2. Confirm which entity the pitch is about before you take a meeting. If a tool offers you "ChatGPT ads" this quarter, ask whether it is Amazon DSP inventory or a direct OpenAI Ads Manager buy. They are different products with different billing and different measurement.
3. Treat any performance figure you are quoted as unverified. A pilot with select U.S. advertisers has not produced published benchmarks. Anyone citing a conversational-ads CPC or ROAS number to an EU seller today is extrapolating.
4. If you already run DSP and sell in the U.S., this is a supply test, so size it like one — a budget you can lose, measured on incrementality, and ring-fenced from the Sponsored Products P&L that pays your bills.
5. Watch your own top-of-funnel keywords for the leading indicator: broad, research-style terms losing volume while your exact-match converters hold or improve. That is the shift arriving in data you already have.
6. Diarise 28–30 September. Amazon Ads holds unBoxed in San Francisco, and product detail — including anything about markets beyond the U.S. — is more likely to land there than in this week's press cycle.

Where we stand on it

Mirox will not be adding a ChatGPT channel. That is not caution about the surface; it is the same position we hold on Walmart, Instacart and Google. Mirox is an Amazon specialist by choice, on the view that depth beats breadth at this stage, and a bidding system that spreads itself across channels it cannot model properly is how sellers end up with a dashboard that reports everything and decides nothing.

It is also a straightforward fit problem. Mirox exists to price a bid against a unit's real margin and show its work — every decision exports with the Bayesian conversion estimate, the alternatives ruled out, and which of the seven safety gates were checked. A surface where the platform controls all delivery decisions and returns aggregated insight is one where that trace cannot be produced. We would rather say so than ship a channel toggle that quietly stops explaining itself. The reasoning behind that rule is in why we built a PPC AI that explains itself.

The one-line version

Amazon Ads announced a ChatGPT advertising partnership on 10 September 2026; it is a DSP-side pilot with select U.S. advertisers, it does not touch EU Sponsored Products, and the right response is to keep bidding to your break-even ceiling while watching your research-intent keywords for the shift that actually reaches you.

See what a full decision trace looks like, or read how conversational shopping is already changing Amazon search.