Amazon is no longer experimenting with AI in advertising. This time, it brought everything under one roof: display, video, audio, targeting, optimization. A single AI agent that decides where, when, and how your ads appear.
It's called Amazon Ads Agent, it's available to all US advertisers, and it comes with numbers that are hard to dismiss: 25% more unique customers reached and CPM reduced by over 10%. Nearly $20 billion in ad revenue in Q2 2026 alone.
But before we get carried away, let's look at what it actually does, what the data shows, and where you need to be careful.
What Amazon Ads Agent actually does
A recent piece on Marketing Dive breaks down what Amazon announced through Kelly MacLean, VP of Amazon Ads. It's a unified platform that merges the DSP (demand-side platform) with the ads console into a single interface.
If you previously managed Sponsored Products, display, video, and audio as separate campaigns, they now run through one instrument. Amazon calls the new format DVA+ (Display, Video, Audio Plus) and plans to roll it out to all advertisers by the end of 2026.
What does the AI do, specifically? It takes an objective ("I want awareness" or "I want conversions"), analyzes audience data from Amazon Marketing Cloud, and decides which combination of formats works best. You set the budget and the goal. Everything else happens automatically.
This is a significant step from what Amazon offered before, where each format had its own workflow, its own settings, and its own reporting. Now everything is centralized, with AI allocating resources in real time.
One component worth noting: Creative Agent. Amazon added an AI module that generates and adapts creatives automatically based on the format. If you have a single set of assets, the AI adapts them for each channel. It's not just budget distribution, it's content production at the format level. Dentsu, one of the largest agency networks, confirmed that internal test results outperformed manually managed campaigns.
The numbers that matter
This is where it gets interesting. Not because Amazon says "AI is good" (everyone says that), but because they bring concrete data from the beta period:
Advertisers who used natural language targeting (conversational AI, where you describe your audience in words rather than technical parameters) reached 25% more unique customers while reducing CPM by over 10%. More reach at lower cost. That's the formula every media buyer is looking for.
Canine Naturals, a dog food brand, ran a full-funnel campaign and generated 41% of sales from entirely new customers over a 206-day period. Brand+ and Performance+ (Amazon's AI optimization tools) delivered an 18-point lift in ad awareness and a 14% reduction in cost per conversion.
These numbers come from Amazon, so take them with appropriate skepticism. But the direction is clear: AI isn't just optimizing what you already have. It's finding audiences you wouldn't have chosen manually.
What this means for advertisers beyond the US
Let's be practical. Amazon Ads isn't as dominant in markets like Romania as Google Ads or Meta Ads. But the trend Amazon is setting matters for every advertising platform, including conversational ads on ChatGPT.
Full-funnel is becoming the default. The separation between "awareness campaigns" and "conversion campaigns" is fading. Platforms are pushing advertisers to let AI allocate budgets across the entire funnel. Google already does this with Performance Max. Amazon now does it with Full-Funnel Campaigns. If you still plan campaigns in isolated silos, you'll lose efficiency.
Media buying work is shifting. AI will take over a significant portion of the operational side: audience targeting, bidding, format distribution. Strategy, messaging, and creative remain yours. But the part where you spend hours on manual optimization rounds is shrinking. It's a trend we explored when we discussed how AI is changing click behavior.
Control is moving. Until now, a good media buyer was someone who knew exactly which bid to place on which keyword. Now, a good media buyer is someone who knows which objective to set, which creatives to prepare, and how to interpret the data AI reports back. The question is no longer "what bid do I set?" but "what do I tell the AI to do?".
This isn't necessarily a bad thing. A media buyer with experience who understands both data and the client's business becomes more valuable, not less. AI eliminates repetitive work, but someone still needs to decide whether the objective is right, whether the creative communicates the right message, and whether the reported data makes sense in context.
Where to be cautious
Automation is tempting, but it has limits. When you let AI decide everything, you lose granularity. You no longer know exactly why an ad performed. Was it the creative? The audience? The time of day? Amazon gives you aggregate results, not the mechanism behind them.
Transparency remains a concern too. When Amazon tells you "AI chose the optimal audience," you don't get details on how it made that decision. For a brand that wants to understand who their customers are, not just get conversions, that can be frustrating. The control you enjoyed with manual campaigns doesn't automatically transfer to the AI world.
And in smaller markets, data volume is critical. AI works well with large conversion volumes. With small budgets and narrow audiences, automated decisions can be less precise. If you sell on Amazon.de with a limited budget, the AI won't have enough signals to learn from.
A pragmatic approach: use automation for high-volume campaigns with enough data, but keep manual control for niche or small-budget campaigns. The two modes don't exclude each other. The combination is what delivers results.
Amazon Ads Agent isn't the future. It's the present. And even if you don't advertise on Amazon, this model of "set the objective, AI executes" is coming to every platform. If you sell on marketplaces integrating AI, prepare to work differently.
The question is no longer whether AI runs ads better than you. The question is what value you add on top of what AI already does.
