on the article · Technology

Amazon’s ad auction had a fake bidder, the FTC says

The FTC says Amazon secretly added a hidden reserve price, turning second-price ad auctions into first-price ones since 2019.

By The Signal · · 5 min read

Amazon's ad auction dashboard, at the centre of the FTC's pricing complaint.
Amazon's ad auction dashboard, at the centre of the FTC's pricing complaint. — on the article

For more than seven years, the Federal Trade Commission alleges, Amazon told advertisers one set of auction rules and ran another. The complaint, filed August 31, 2026 by the FTC and 22 state attorneys general, says Amazon charged Sponsored Products advertisers their own full winning bid close to 80 percent of the time — in an auction it marketed as never charging more than one cent above the next-highest bid.

Close to 80 percent, and where that number comes from

The 80 percent figure is the spine of the case. It comes from the complaint itself, which the FTC says draws on Amazon's internal auction data and documents. The claim is narrow and specific: in a "second-price" auction, the winner should pay one cent more than the second-highest bid, no matter how high their own bid went. If that mechanism worked as advertised, an advertiser's own bid amount would rarely equal what they actually paid — the market, not their own number, would set the price. The FTC says that stopped being true for most Sponsored Products auctions after 2019, when the winning price and the winning bid converged almost every time.

That convergence is the tell. A functioning second-price auction produces a gap between what you offered and what you paid. Collapse that gap to zero in four out of five auctions and you no longer have a second-price system — you have a first-price one wearing a second-price label.

How the "soft reserve price" actually worked

The mechanism the FTC describes has a name inside Amazon, according to the complaint: a "soft reserve price." Rather than compare a winning bid only to the next real bidder, Amazon inserted a threshold — internally described in one document as an "invented auction participant" — that acted like a bidder who was never actually competing. If the real second-highest bid fell below that threshold, the threshold itself set the price instead. Think of a live auctioneer who quietly has a floor price in mind and, when real bids don't reach it, announces a bid from someone standing in the back who isn't actually there. Buyers keep raising their offers because they believe they're competing against each other, when part of what they're competing against is a number the house set. The FTC calls this a shill bid: a price manufactured internally rather than discovered through competition.

The complaint says Amazon introduced this change in 2019 without telling advertisers, and continued describing its system publicly, in training materials, and through its sales staff as a standard generalized second-price (GSP) auction — the same auction format used across digital advertising, from search to social media, which advertisers rely on to calibrate how aggressively they bid.

Why the auction type changes how a business bids at all

The distinction between first-price and second-price auctions isn't cosmetic; it changes behavior. In a first-price auction, where you pay exactly what you bid, rational bidders shade their bids downward over repeated rounds — offering less than they think something is worth, because overbidding is a direct loss. In a genuine second-price auction, that caution is unnecessary: you can bid closer to your true value because the mechanism guarantees you'll only pay the minimum needed to win. The FTC's argument is that Amazon got the advantages of both: advertisers bid like it was second-price — aggressively, near their real valuations — while Amazon in practice collected first-price money close to 80 percent of the time. More than 500,000 small and medium-sized businesses were told the version of the story that kept their bids high.

The number that makes tens of billions plausible

Amazon reported more than $68 billion in advertising revenue last year. The FTC's complaint alleges the surcharge practice affected more than 1 million brands and sellers over seven-plus years and "likely extracted tens of billions of dollars" in aggregate. Set against a single year's $68 billion top line, a cumulative multi-year overcharge in the tens of billions is not an outlandish claim on its face — it implies the alleged surcharge added a modest but persistent percentage to auction prices, compounded across years and across a business with over a million paying customers. The FTC's own framing leans on scale: Chairman Andrew Ferguson called the effect "staggering" and said the costs were "largely passed on to American consumers" — the ad markup, in other words, functions as an invisible line item in retail prices, absorbed by sellers and passed through to shoppers who never see an ad auction.

The counter-argument, stated plainly

Amazon's response, published in a company blog post, is that the FTC "fundamentally misunderstands how advertisers operate." Its position: the platform runs billions of bids across many placements and ad formats simultaneously, so prices for what looks like "the same" keyword slot will naturally vary by context, timing, and competition — and that advertisers are properly informed about how pricing works. This is a real point about complexity. A single keyword auction is not one static event; it recurs continuously, at different times of day, against different competitors, for different ad formats (Sponsored Products, Sponsored Brands, Display). Variation alone would not prove wrongdoing. The complaint's answer to that is not the variation — it's the internal documentation. The FTC is not alleging that prices moved around; it is alleging that Amazon's own internal materials describe a hidden threshold with a name, a rationale, and a timeline for when it was added. Variance explains noise. A named internal mechanism, introduced in a specific year and never disclosed, is a different kind of evidence.

What the case does next, and the constraint nobody's litigating

The suit joins Alaska, Arizona, California, Colorado, Florida, Idaho, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maryland, Nebraska, New Jersey, New York, North Carolina, Oklahoma, Pennsylvania, Rhode Island, South Carolina, Vermont, and Washington as co-plaintiffs alongside the FTC — a broad multistate coalition that increases the odds of a coordinated settlement rather than a single-state resolution, and raises the ceiling on potential penalties and disgorgement if the allegations hold up. But the case, as filed, is about disclosure and mechanism — what Amazon told advertisers versus what its systems did. It does not, on its own, restructure how ad auctions work industry-wide, and it does not touch the harder question underneath: advertisers have no independent way to audit an ad platform's auction logic. They see a bid box and a bill. Everything in between — the reserve price, the ranking formula, the "invented participant" — is proprietary. Whatever this case resolves about Amazon's specific conduct, it will not give advertisers, on Amazon or anywhere else, a way to verify the next platform's auction from the outside. That verification gap outlasts any one lawsuit.