Amazon Just Put ChatGPT Ads Inside the Media-Buying Stack

Amazon advertisers can now buy ChatGPT ads without adding a separate media-buying platform. For a small marketing team, that removes one reason to ignore the channel, but it does not answer the harder question: whether the ads produce customers.

Amazon announced the U.S. pilot on September 10. Select advertisers can extend campaigns from Amazon DSP into ChatGPT while Amazon helps set up and improve the buy. OpenAI still decides when an ad appears and where it is placed.

That division of labor matters. Amazon brings the advertiser relationships, campaign workflow and audience signals. OpenAI keeps control of the conversation and the inventory. ChatGPT advertising just moved closer to the media plan a buyer already runs, without becoming an ordinary display placement.

What Amazon opened

The pilot is limited to select U.S. advertisers. Marketing Dive reported that the inventory is available as a managed service through Amazon DSP on cost-per-click and cost-per-thousand-impression pricing. Text and image units appear beneath ChatGPT responses with sponsored labels. Product-feed ads can also generate assets from a catalog.

Amazon’s pitch is access to people while they are comparing options and making decisions. Chris Conetta, director of omnichannel supply at Amazon DSP, put it this way:

“With access to ChatGPT Ads, advertisers can now extend their campaigns further into where their customers are actively spending time in conversational chat platforms.”

Delta Vacations is among the first named testers. Its president, Katrin Koenig, said the company can use the collaboration to “inform how and when Delta Vacations appear within ChatGPT Ads experiences.”

The wording is careful. Neither company says Amazon chooses the exact prompt or answer beside which an ad runs. Marketing Dive says OpenAI controls delivery through its own systems, while advertisers receive aggregated measures including impressions, clicks, cost per result, CPM and CPC.

What changed for a marketing team

Until this deal, testing ChatGPT ads meant treating OpenAI as another emerging platform with its own buying process, reporting and learning curve. Amazon has now put the inventory beside media that advertisers already buy through its DSP.

That makes a test operationally easier. It also changes who can put ChatGPT on the plan. The buyer no longer needs a direct OpenAI relationship, and a brand that does not sell on Amazon can still use Amazon DSP. Amazon’s own documentation says nonendemic advertisers can use the platform across properties beyond its store.

The bigger strategic change is distribution. OpenAI said ChatGPT ads reached a $1 billion annualized run rate before publishing channel-level conversion benchmarks. Amazon can now introduce that inventory to a much larger base of established advertisers. Ease of purchase may push spending ahead of evidence even faster, especially after OpenAI removed the platform’s hard daily budget cap.

The performance problem did not move

Amazon’s announcement contains no conversion rate, acquisition cost or return on ad spend. Delta Vacations says it is helping define the approach, not that it has measured a result. This is a pilot, not a benchmark.

Digiday’s reporting adds the useful brake. Nate Elliott, principal analyst of AI at eMarketer, said:

“They’re still trying to build out even many of the basics of a functional ad sales operation, including their team, technology, vendor partnerships, ad formats, and pricing.”

That caveat gets more important when a familiar buying interface makes a young channel look mature. The Amazon DSP can make the transaction familiar. It cannot make ChatGPT traffic comparable to search, social or retail media before enough campaigns disclose outcomes.

There is also an attribution problem inside the product itself. A person may see an ad while researching in ChatGPT, continue the conversation, leave later and buy elsewhere. The metrics Amazon lists are useful for delivery. They do not by themselves prove incrementality or explain what the surrounding answer contributed.

How to treat the pilot

Do not add ChatGPT merely because it appeared inside a platform your team already trusts. Put it through the same approval gate as a new channel.

First, write the hypothesis. Name the audience, the decision they are making in ChatGPT and the action that would count as success. “Try AI ads” is not a hypothesis.

Second, isolate the test. Use distinct campaign tags and a landing experience you can measure outside the DSP. Compare the result with the next dollar in your current channel, not with doing nothing.

Third, ask Amazon what is actually available to your account. The public announcement does not name minimum spend, eligibility rules, inventory volume or a timeline beyond the select U.S. pilot. Those unknowns belong in the buying decision, not in a footnote after launch.

Amazon has made ChatGPT ads easier to buy. That is real news for media teams. It has not made them easier to evaluate, and the distance between those two facts is where an experimental budget can quietly become a recurring line item.