Amazon Ads Inside ChatGPT: What the OpenAI Pilot Means for Amazon Sellers
Shoppers who ask ChatGPT what to buy can now see your product in the answer, bought through the Amazon Ads account you already run. On September 10, 2026, Amazon Ads announced a partnership with OpenAI that lets select US advertisers extend Amazon DSP campaigns into ChatGPT, with Delta Vacations named among the first brands testing it.
Most of the coverage has treated this as an ad-tech story about two large companies. For an Amazon seller the story is more useful than that. Roughly 900 million people use ChatGPT every week, a growing share of them to research purchases, and until now the only way to advertise there was through OpenAI's own Ads Manager, which measures conversions with a pixel you cannot place on Amazon.com. This pilot is the first route where the click lands on your listing and Amazon's own attribution sits in the loop. It is also a managed, invite-only pilot with reporting Amazon has not fully described, so the case is for a bounded test, not a new budget line.
Quick Answer
- What launched: a US pilot, announced September 10, 2026, that lets select Amazon Ads advertisers extend Amazon DSP campaigns into ChatGPT. Ads appear as labeled text or image units below ChatGPT's answers.
- How you buy it: through Amazon DSP as a managed service. Amazon's team sets up and runs the campaign, priced on CPC or CPM, and OpenAI's ad system decides which ads appear and when. There is no self-serve option yet and no announced date for one.
- The opportunity for sellers: shoppers are already asking ChatGPT for product recommendations, product feed ads generate the creative from your catalog, and the click goes to Amazon, where a purchase can be attributed without a pixel you cannot install.
- The risk: Amazon has published no detail on how a ChatGPT click is attributed or deduplicated against the rest of your DSP reporting, OpenAI controls delivery, and the pilot is invite-only and US-only.
- Our read: if you run Amazon DSP already, or your Amazon Ads partner does, ask for a small, bounded test now while the surface is new. Hold the rest of the account still so you can read the result. Wait for a published measurement model before you treat ChatGPT as a scalable channel.
What Amazon and OpenAI Announced
The mechanics are narrower than "Amazon ads are now in ChatGPT" suggests. Amazon's announcement describes the arrangement as a way for advertisers to expand existing Amazon Ads campaigns with a conversational ad experience on ChatGPT, tested by select US advertisers as a pilot. Chris Conetta, Director of Omnichannel Supply at Amazon DSP, called conversational ads "the fastest growing engagement opportunity for brands to reach new and existing audiences."
The operating details came from an Amazon spokesperson to Marketing Dive and CNBC rather than from the announcement itself:
- Inventory is bought through Amazon DSP and offered as a managed service, so Amazon's team helps set up and optimize the campaign.
- OpenAI controls all ad delivery decisions through its own systems.
- Ads surface as text and image units beneath ChatGPT's organic responses, clearly labeled as sponsored, on ChatGPT's free and Go tiers.
- Pricing runs on standard CPC and CPM models.
- Amazon is also offering product feed ads that generate assets from an advertiser's product catalog.
- Amazon's representatives offer "context hints" to match ads with relevant topics, conversations, and keywords, since OpenAI's system matches on conversational context rather than a keyword auction.
None of that is a knock on the announcement. It is the normal shape of a pilot: the inventory arrives first and the measurement layer that proves it out arrives later.
Why This Matters for an Amazon Seller
Three things make this worth a seller's attention, rather than only a media buyer's.
The demand is already there, and Amazon is already buying it
Amazon itself ran sponsored Prime Day placements inside ChatGPT in June 2026, sending shoppers who asked for the best deal on a product straight to Amazon.com, as Modern Retail reported. Amazon spent money to pull ChatGPT research traffic onto its own store before it offered the same route to sellers. That is as clear a signal as you will get that the surface converts to Amazon visits.
The click lands somewhere Amazon can measure
OpenAI has built real conversion measurement into ChatGPT Ads over 2026: a site pixel, a Conversions API, attribution windows, and, since August, conversion-optimized bidding for product feed campaigns. All of it assumes you control the destination page. An Amazon seller does not. You cannot put OpenAI's pixel on your Amazon listing, so a direct buy through OpenAI's Ads Manager sends traffic you can see and purchases you cannot. Buying through Amazon DSP flips that: the click goes to Amazon, where DSP already attributes off-Amazon traffic to detail page views and purchases. Amazon has not said how ChatGPT clicks appear in that reporting, which is the open question below, but the route is the right one for a seller whose sales happen on Amazon.
Product feed ads keep the creative lift small
A managed pilot that generates the ad from your catalog means the work on your side is choosing the ASINs and making sure their detail pages can carry the traffic. The same listing quality that decides whether you show up in Alexa for Shopping recommendations is what a ChatGPT shopper lands on. Nonendemic brands that use Amazon DSP without selling on Amazon get the same route, which is part of why Amazon built it.
What the Reporting Tells You, and What It Does Not
Advertisers in the pilot receive aggregated performance data covering impressions, clicks, cost per result, CPM, and CPC, according to TechWyse's account of the announcement. That is a real reporting surface, and "cost per result" implies some result is being counted. What is not published is anything a seller would need to trust the number:
- Whether OpenAI's conversion tools apply. The pixel, Conversions API, and oCPC bidding exist for OpenAI's direct advertisers. Amazon has not said whether any of that flows through the managed DSP route, or whether the result Amazon reports is Amazon's own attribution on the destination page.
- How a ChatGPT click is deduplicated. If a shopper clicks a ChatGPT ad on Tuesday and a Sponsored Products ad on Thursday, nobody has said which touchpoint gets the purchase in your reporting. This is the same question every DSP extension onto third-party inventory raises, and buyers running LinkedIn CTV ads inside Amazon DSP know how long it can take to get an answer.
- What OpenAI is optimizing for. OpenAI's system decides when and where an ad appears. Amazon's optimization logic does not run inside ChatGPT, so performance depends on a delivery engine neither you nor your Amazon rep can see.
- Who gets in. The pilot is US-only and limited to select advertisers. Amazon has not published selection criteria, a self-serve timeline, or an expansion date.
A seller who cannot verify the result should treat the number as directional and lean on the one thing Amazon does report cleanly: purchases on Amazon during the test window, read against a baseline you set before the test starts.
ChatGPT Ads Next to Amazon's Own Conversational Placements
Amazon has been building conversational ad surfaces on its own store for a while. Sponsored prompts inside Alexa for Shopping, formerly Rufus, run inside a surface Amazon owns, instruments, and reports on with its own attribution stack. ChatGPT is different: Amazon is a buyer there, using the DSP relationship as the entry point. A seller treating "conversational ads" as one category will misjudge how much to trust the reporting on either one.
The distinction matters for sequencing. A brand already running sponsored prompts has a baseline for how conversational placements perform against search and display in the same account. That baseline does not transfer to ChatGPT: different platform, different shopper behavior, different measurement, and a managed service standing between you and the levers.
How to Run a Bounded Test
The pilot is worth testing precisely because it is early and thinly contested. The way to test it without fooling yourself is to make everything else in the account boring for the duration.
- Pick a small set of ASINs that can carry the traffic. Strong detail pages, healthy margin, and inventory cover for a spike. Shoppers arriving from a ChatGPT recommendation are comparing, so the listing has to answer the question they asked.
- Use the product feed. Let the creative come from the catalog. Spend your effort on which products are in the feed, not on ad copy.
- Set a budget you can afford to learn from and a fixed window, four to six weeks, so you are reading a complete period rather than a partial one.
- Write the baseline down before launch. Detail page views, purchases, and new-to-brand rate for the test ASINs, plus pacing and spend across the rest of your DSP and sponsored campaigns.
- Decide what "worked" means in advance. A lift in Amazon-attributed purchases on the test ASINs against that baseline, not the pilot's own cost-per-result figure, which you cannot yet verify.
- Hold the rest of the account still. No budget reallocations, bid strategy changes, or new campaigns on the test ASINs during the window. Every change you do make gets logged with a reason.
Where Qinetix Fits
Step six is the one that breaks in practice. A live Amazon account does not sit still for six weeks: bids move, budgets pace, someone launches a campaign, and by week four nobody can say whether the lift on the test ASINs came from ChatGPT or from a bid change on the same products. That is the specific job Qinetix, the ads and pricing execution layer of Trellis, does during a test like this. It runs your Sponsored ads and DSP bids inside zones you set, with a floor and a ceiling, so the rest of the account holds the pattern you chose instead of drifting, and every change lands in a per-campaign history log with the rule that caused it. When the test window closes, the question of what else moved has a written answer, the same visibility that matters during any automated cross-format campaign.
What Qinetix does not do, and nothing outside Amazon and OpenAI can do right now, is see inside ChatGPT's delivery decisions or deduplicate a ChatGPT click against an Amazon touchpoint when the two platforms have not published how they do it themselves. Any vendor claiming otherwise is guessing. Qinetix keeps your side of the ledger clean so the ChatGPT line is the only unknown in it.
Conclusion
The pilot is real, the route is the right one for a seller whose sales happen on Amazon, and the measurement question is unanswered rather than answered and unpublished. That combination argues for a small test now, run against a baseline you wrote down, on a set of listings ready for a comparison shopper. It argues against making ChatGPT a scalable line item until Amazon publishes how a ChatGPT click is attributed and deduplicated, the same wait-and-test posture worth applying to any of the AI-driven changes Amazon has rolled out across ads in 2026. If you are not on Amazon DSP yet, our guide to the Amazon Demand-Side Platform covers what a seat involves, since a DSP relationship is the door to this pilot.
Frequently Asked Questions
No. It's managed service only. Amazon's team sets up and optimizes the campaign for the advertiser through Amazon DSP, with product feed ads that generate creative from your catalog. There's no self-serve console yet and no announced date for one.
Delta Vacations is the first named test advertiser. Amazon has not published a full list or confirmed how advertisers get selected for the US pilot.
Partly. OpenAI's ChatGPT Ads platform offers a pixel, a Conversions API, and attribution windows, but those assume you control the landing page, and a seller can't place a pixel on Amazon.com. Through the DSP route the click lands on Amazon, where Amazon already attributes off-Amazon traffic to detail page views and purchases. Amazon has not published how ChatGPT clicks appear in that reporting or how they're deduplicated against your other campaigns, so read purchases on your test ASINs against a baseline you set before launch.
Not yet. The pilot is limited to the US as of September 2026, with no confirmed expansion timeline.
If you're already running meaningful Amazon DSP volume and can afford a small, bounded test, testing now while the surface is new and lightly contested is defensible. If your team needs clean attribution to justify spend internally, waiting for a published measurement model is the more conservative call.
No. Sponsored prompts run inside Amazon's own surface with Amazon's attribution stack, and eligible Sponsored Products and Sponsored Brands campaigns extend into them automatically. The ChatGPT pilot is Amazon buying inventory on a platform it doesn't own, through a managed DSP relationship, with reporting Amazon hasn't fully described.
CPC and CPM, the same buying models Amazon DSP already uses elsewhere. No published pricing floor or benchmark specific to the ChatGPT placement yet.
For this route, yes. The pilot is bought through Amazon DSP as a managed service, so you need a DSP relationship, either directly or through an Amazon Ads partner. The alternative is OpenAI's own Ads Manager, which is self-serve but can't attribute purchases that happen on Amazon.com.