# Amazon's Ad Auction Ran on a Bidder That Did Not Exist
On September 1, the Federal Trade Commission and attorneys general from 22 states sued Amazon in federal court in Seattle, alleging that the company's advertising auctions, long described to advertisers as "second-price," quietly stopped working that way, at a cost the government puts at more than $20 billion since 2019. The complaint says the practice touched roughly 1.2 million advertising customers, more than half a million of them small and medium-sized businesses. Amazon answered within hours, calling the lawsuit misguided and saying the agency "fundamentally misunderstands" how ad auctions work. The case will be decided slowly, but its shape is already visible: it is a dispute over a price that was never printed.
The name of the auction was doing the pricing
The starting point is a mechanism most advertisers only half see. In a second-price auction, the highest bidder wins but pays just above the second-highest bid, one cent more in Amazon's design. A bidder who offers $200 against a $190 rival pays $190.01. Amazon adopted the design in 2012 and marketed it as a fairness feature: bid what the placement is truly worth to you, because the market, not your nerves, decides what you pay. The incentive to shade bids disappears. The price is anchored to what someone else was willing to pay, not to your own guess.
The FTC's complaint says that beginning around 2018 or 2019, the mechanism drifted away from its name. By the complaint's arithmetic, Sponsored Products advertisers paid their full winning bid in roughly 30 to 40 percent of auctions in 2021, about 70 percent in 2022, and around 80 percent by 2024. The gap between the promise and the price was widest on the busiest shopping days, Prime Day and Black Friday among them. A second-price auction that charges the first price is simply a first-price auction, and the difference between the two is where the money in this case lives.
A bidder that did not exist
The mechanism the complaint describes is a price floor dressed as a participant. Internal documents, the FTC says, refer to an "invented auction participant," a fictitious bidder carrying Amazon's own valuation of an ad slot, an arrangement sometimes described as a soft reserve price. A real second-highest bid is a fact about the market. A manufactured one is a decision by the auctioneer about what the slot is worth. The effect on the winner is similar either way: the price rises. The difference is who can see why.
The complaint alleges that the phantom bidder was never disclosed, so advertisers bid into a game whose second player they could not observe. Employees saw the tension, according to the FTC: one internal memo acknowledged that the undisclosed reserve prices would lift short-term revenue but would, in the document's words, "hurt us in the long run." The complaint's framing is that the change followed plain arithmetic, an auction business that was not generating the revenue its operator thought it should.
The defense is measured in a different unit
Amazon's response is a list of numbers that answer a different question. Average cost per click for Sponsored Products stayed flat after inflation from 2019 through 2024, the company says. Average winning bids fell by half. Its relevance-based auction design saved advertisers more than $8 billion between 2021 and 2025, and no advertiser ever paid above its own stated maximum bid. Amazon also argues the FTC cites no evidence that consumer prices rose. Read that way, the check still cleared: advertisers bought clicks at a price that did not, on average, climb.
There is a deeper point inside Amazon's defense, and it deserves to be stated at full strength. The company has long argued that its ads are priced by more than bids: a relevance score, built on how likely a shopper is to buy, is woven into every auction. In that view, the auction was never a pure second-price market, and the promise advertisers received was never the mechanical one the complaint reconstructs. The FTC, Amazon says, has taken a complex, relevance-driven system and re-described it as a rigged coin toss. Even a sympathetic reader can see why the agency's case is not trivial to prove: it must show not only that undisclosed mechanics raised some prices, but that advertisers would have paid meaningfully less under the disclosed one, and that the gap persisted at scale. That is a counterfactual inside a counterfactual.
The problem for the narrative is that both sets of arithmetic can be true at once, because the two sides are measuring different quantities. The government's overcharge is a counterfactual: what would the winning bidder have paid in a genuine second-price auction? That number is never observed anywhere. It has to be reconstructed from internal documents and the frequency of full-bid pricing. Amazon's defense is built on observed prices: what advertisers actually paid, on average, over time. A market can charge more than the counterfactual while its average price holds flat, if the counterfactual would have fallen, or if the mix of auctions shifted. Neither side can settle the argument with its own ledger, because the dispute is about a price that does not exist on any receipt.
The people who paid were not the ones who set the bids
The asymmetry of the marketplace matters as much as the arithmetic. Amazon saw every bid. An advertiser saw exactly one: its own. In a second-price auction, the auctioneer's view of the losing bid is the entire machinery, which is why the second bid has to be real for the promise to mean anything. The allegation is that the machinery ran the other way, holding prices up rather than discovering them.
The people on the other side were, by and large, not sophisticated media buyers. The complaint counts more than half a million small and medium-sized businesses among the affected advertisers, sellers who treat advertising as a line item and an auction as a black box. The state attorneys general make the pass-through point directly. New York's Letitia James said consumers are "likely paying more for everything from groceries to electronics" because ad costs end up in retail prices. FTC Chairman Andrew Ferguson said advertisers were "misled into paying significantly higher prices," and that the higher costs were largely passed on to American consumers. The striking part is the silence the complaint implies: a bidder cannot notice a price being pushed up by a bidder that does not exist.
The third act of a longer argument
This is the third major FTC action against Amazon, and it lands at a crowded moment. The company paid $2.5 billion last year to settle allegations over Prime subscription practices, and a separate monopolization trial over its retail marketplace is scheduled for early next year. Texas has filed its own case over advertising pricing. The advertising business explains the stakes: it is the company's margin engine, the division whose profits fund everything else, and the integrity of its pricing mechanism is the asset every other business borrows against. Markets registered the suit mildly, with Amazon shares down roughly 2.5 percent on the day, a small move for a company of this size and its own quiet statement about the odds. The remedies sought, civil penalties, restitution, and an injunction, would take years either way. The allegations are unproven and denied in full.
The auction's real product was trust
Step back and the case is about a promise embedded in a name. "Second-price" was never merely a technical detail. It was a transparency contract: bid what the placement is worth to you, and the market will set the price. Advertisers built budgets, automated bidding strategies, and years of accumulated trust on that promise. The complaint's theory is that the promise became decoration while the pricing did something else. Amazon's theory is that the decoration was beside the point, that what advertisers bought was performance at a market price, and that the performance was delivered.
The two positions cannot both be right about the history, but they can both be honestly held, which is what makes the case hard and worth watching. Whatever the courts conclude, the episode clarifies what an advertising auction actually sells. It sells reach, and it sells measurement. Underneath both, it sells the belief that the price is what the price would have been. A bidder that does not exist is the fastest way to spend that belief.
The practical lesson for the people who buy ads arrives before any verdict. Pricing rules are not decorations. When a platform describes how its auction works, that description is part of the product, the way a nutritional label is part of food, and the complaint's central allegation is that the label stopped matching the contents. Advertisers who read this case closely will not wait for the outcome to change their behavior. They will ask platforms the question the lawsuit makes unavoidable: what rule actually sets my price, and who can see the second bid? The fact that the question sounds naive is the point. The most sophisticated ad market in the world may now spend years litigating what, exactly, its own auction was. This is general analysis and not legal advice.
---
Primary sources
- Engadget's report on the filing supplied the hidden-surcharge language, the $20 billion figure, the 1.5 million pages reviewed across six years, the employee memo and its "hurt us in the long run" language, and Amazon's response, including the "fundamentally misunderstands" quote.
- Alaska Beacon's coverage of the complaint supplied the 1.2 million advertiser count, the small-business figure, the 2012 second-price design, the full-bid frequency progression across 2021 through 2024, the invented auction participant and soft reserve price mechanism, and the Prime Day and Black Friday detail.
- NBC Boston's report supplied the 22-state roster and Ferguson's statement that advertisers were "misled into paying significantly higher prices" with the costs largely passed on to consumers.
- Benzinga's market report supplied Amazon's flat cost-per-click claim, the halving of average winning bids, the $8 billion savings figure, the statement that no advertiser paid above its maximum bid, and the roughly 2.5 percent share decline.
- The Columbian's report supplied the James and Bonta statements, including the "groceries to electronics" quote, and the broader regulatory background of the Prime settlement, the upcoming retail trial, and the Texas case.
The reading of the dispute as one over an unobservable counterfactual price, the framing of the phantom bidder as a price floor dressed as a participant, and the argument that the auction's real product is trust in the pricing rule reflect general analysis synthesized from these sources rather than any single one.