Amazon wants advertisers to hand more campaign decisions to AI. The problem is that some advertisers and regulators are still asking a simpler question: can they see how Amazon’s existing ad decisions are made?
That tension is becoming harder to ignore. In its reporting on Amazon’s auction trial, Digiday details newly unredacted FTC allegations that Amazon used undisclosed reserve-price mechanisms in Sponsored Products auctions. Amazon disputes the allegations, and the case remains ongoing.
The black box comes before the AI
According to the allegations, Amazon presented its auctions as generalized second-price systems, where advertisers would generally pay based on competing bids rather than their maximum bid. Regulators allege that reserve prices pushed charges higher without corresponding competition, making the system operate more like a first-price auction.
The figures are significant. The FTC alleges that Amazon imposed surcharges of up to 50% on ordinary shopping days and 250% during major retail events, generating an additional $87 million during the 2022 holiday season. It also alleges that average cost per click rose 20% during the 2021 holiday season, with some advertisers seeing increases of more than 200%.
Amazon’s position is that reserve pricing can reflect the value of an advertising placement and is not inherently unlawful. That dispute will be settled through the legal process. The immediate issue for advertisers is more practical: they can see what they paid, but may not be able to independently establish why they paid it.
Automation raises the stakes
The allegations arrive as Amazon expands the amount of advertising work it wants to automate. Its Ads Agent and Full-Funnel Campaigns products let advertisers set objectives, products, creative, and budgets while Amazon’s systems handle targeting, allocation, and optimization.
Amazon says beta testing produced up to 67% higher return on long-term ad spend and 29% lower costs per new-to-brand customer compared with manually configured campaigns. Those are company-reported figures, but they describe the bargain Amazon is offering: less manual control in exchange for promised performance.
That bargain becomes harder to accept when the platform’s underlying pricing mechanics are already under scrutiny. The FTC case is about Sponsored Products auctions, not proof that Amazon’s newer AI tools are behaving improperly. But it exposes the trust requirement those tools will face. If advertisers cannot inspect how money is priced, allocated, or optimized, better automation does not remove the black box. It makes the black box responsible for more of the budget.
For Amazon, the strategic consequence is clear. AI adoption in advertising will depend on more than performance claims. Advertisers will also need enough visibility into the system to believe that efficiency is being created for them, rather than simply being reported by the platform controlling the transaction.