The retail giant's ChatGPT ad deal isn't about AI. It's about making sure every advertising dollar passes through its hands on the way to wherever it's going.

From 10 September, brands that advertise through Amazon Ads can buy ad placements inside the ChatGPT app. That sentence sounds like the headline, but it's actually the least interesting part of this story (which is saying something, given that it involves two of the most aggressively expanding companies in tech quietly deciding to share a wallet).

The mechanism is Amazon's demand-side platform—the same console advertisers already use to buy inventory elsewhere—which now connects to ad space OpenAI is opening up inside ChatGPT. In practice, an advertiser can plan and buy ChatGPT placements using Amazon's shopping and browsing data to decide who sees them. It's a pilot rather than a general launch: limited to a select group of US advertisers, with Delta Vacations among the first to take part. This is a supply arrangement—OpenAI provides the ad space, Amazon provides the advertisers and the buying pipe—not a merger, not an equity deal, and not any change to how ChatGPT works for ordinary users.

If you're an Amazon seller reading this and wondering whether you should be panic-learning a new ad console, the answer is no. The entry costs are steep, the measurement is thin, and most ecommerce operators won't touch AI-surface advertising for a while yet.

But the structural signal underneath? That's the part worth your attention. Because Amazon isn't trying to build a chatbot. It isn't competing to own the conversational surface where consumers ask what to buy. It's doing something far less glamorous and far more consequential: positioning itself as the pipe that every advertising budget flows through, regardless of who owns the screen at the other end.

ChatGPT is simply the newest screen on the list.

To understand why this matters, you need to zoom out from the ChatGPT announcement and look at the pattern. Over roughly the past eighteen months, Amazon has moved its advertising business into streaming services, premium audio, sports broadcasts, and now a leading AI assistant. Different surfaces, different audiences, different content. Same role for Amazon every single time.

That role runs through something called a demand-side platform, or DSP—the tool advertisers use to buy ad space across many different places from a single console, rather than negotiating with each publisher separately. (Think of it as the universal remote control for ad spending, except this remote also knows what you had for breakfast and how long you spent comparing protein powders last Tuesday.)

Amazon's DSP now reaches into premium video, audio, and conversational AI. The company doesn't need to own any of those environments to profit from them. It just needs to own the buying pipe. And the more surfaces that get routed through that pipe, the harder it becomes for advertisers to justify using anyone else's.

This isn't an AI strategy. It's an aggregation strategy. The AI surface just happens to be the latest thing being fed through it.

Aggregation only works if advertisers have a reason to route their spending through one party rather than going direct. Amazon's pitch has two parts, and both are disarmingly simple.

Part one: price. By several accounts, Amazon's DSP undercuts competitors on fees—sometimes dramatically. On certain reserved deals, Amazon has reportedly charged close to zero against the fifteen-to-twenty per cent that independent platforms and Google have historically taken. Those figures come from a limited set of reports and should be read as directional rather than gospel, but the direction is the point. If two systems can buy the same inventory and one charges materially less, budget follows gravity.

Part two: data. Because Amazon is a marketplace as well as an advertising business, it holds first-party shopping and browsing data that no pure ad platform can match. Amazon knows what people search for, what they compare, what they put in their basket and then agonise over for three days before either buying or abandoning in a fit of financial responsibility. Paired with cheap access to premium inventory, the pitch writes itself: buying through Amazon costs less and targets better than buying anywhere else.

So far, that argument has largely held. Amazon's ad revenue has continued to grow, and some of that growth appears to have come directly at the expense of independent buying platforms. Which is roughly the advertising equivalent of watching someone eat your lunch while charging your other friends less for the same meal.

There is a timing problem here that the celebratory framing tends to skip over with impressive grace.

Amazon is extending its reach into a shiny new premium environment at the exact moment that the Federal Trade Commission has an active lawsuit alleging the company misled advertisers into overpaying for inventory. The specifics will play out in court rather than in newsletter commentary, and the allegations are contested. But the juxtaposition is hard to ignore.

The pitch: Amazon offers the cheapest, best-targeted way to buy ad inventory. The lawsuit: in at least some circumstances, advertisers may have been charged more than they should have been. Both things are under discussion simultaneously without either being resolved, which is precisely the kind of discomfort that makes legal teams reach for the antacids.

What's notable is how little visible resistance the lawsuit has produced from buyers. But that's not really surprising. The recurring lesson of the past two decades of digital advertising is that a platform delivering measurable performance can absorb a remarkable amount of concern about how it operates before advertisers actually change their behaviour. Performance buys forgiveness. The ChatGPT deal is a bet that it will continue to.

The arrangement isn't one-directional. OpenAI has been vocal about wanting to build an advertising business, and it's been assembling pieces at speed—cutting deals with agency groups, connecting commerce media firms, generally doing the things you do when you want a hundred billion dollars in annual ad revenue by 2030.

That target, by the way, would require sustained compound growth above two hundred per cent annually from here. Which is the kind of number that looks ambitious on a slide and physically improbable on a spreadsheet.

Amazon would arguably be the most significant demand source to route into ChatGPT if the pilot becomes permanent. But access to Amazon's advertisers wouldn't fix OpenAI's harder problems. Measurement is still thin—advertisers often can't see clearly whether an ad inside a conversation actually led to a purchase. The underlying tech is described as unpolished. The range of available ad formats is narrow.

One analyst covering AI advertising has argued that even the upcoming holiday quarter would be too generous to treat as a real stress test, on the grounds that OpenAI is still building the basic components of a functioning ad sales operation. On that view, the question isn't whether this quarter goes well. It's whether the foundations are solid enough to matter two years from now. Which is a less exciting framing, but probably a more useful one.

The honest starting point: this is not a channel most Amazon sellers or mid-market brands can buy into today. Reacting to it as an immediate media-buying decision would be premature. The value is in what it signals about where things are heading, and there are three things worth filing away.

Concentration is deepening. The party increasingly positioned to route ad budgets into the newest surfaces is the same party that owns the marketplace many sellers depend on, runs the retail media those sellers already pay for, and competes with some of them directly through its own product lines. None of that is new. But each additional surface routed through Amazon's pipe makes the dependency a little deeper and a little harder to unwind. That's not a reason for alarm. It's a reason to price concentration into longer-term planning rather than treating each channel as independent.

Measurement will decide the market. Whoever standardises how performance is counted on AI surfaces will effectively set the terms for everything built on top of them. The theoretical appeal of conversational advertising rests on the richness of intent people reveal when they talk through a decision in detail. That appeal only converts into real money if the intent signal survives the shift from organic conversation to commercially influenced one—and if it can be measured credibly. Both remain open questions.

Watch the access threshold, not the pilot. The most practical thing for operators to track is whether AI-surface inventory eventually becomes available at lower minimums through the same console sellers already use for existing campaigns. If it does, the question stops being structural and becomes operational—and the operators who understood the shape of the market early will make better decisions than those meeting it cold.

The clean version of this story is that AI advertising has arrived and a new channel is opening. The more useful version is quieter, and considerably less tidy.

A single company is working to make itself the point through which ad budgets flow onto surfaces it doesn't own, at the exact moment that both the fairness of its pricing and the readiness of its newest partner remain genuinely unsettled. Whether that position holds—and what it costs the businesses that come to depend on it—will be worked out over the next few years rather than the next quarter.

In the meantime, the smartest move is the boring one: understand the structure, watch the access thresholds, and resist the urge to treat every new deal as either a crisis or an opportunity. Sometimes it's just a pipe getting longer.

📧 Inbox Zero - Clean Up Your Inbox In Minutes

🎨 Sesame - A Creative Tool Purpose-built for Brand Expression

🌐 Anyword AI - Instantly analyze every piece of content you've ever published, so you know which messaging works best on your website, socials, and email channels.

⌨️ Quicky AI - Use AI on any website instantly just by selecting text

😍It was great, all I needed

🙂Good, not great

🙄It sucked