Amazon has now explained its own ad auction to the public, and the explanation is the FTC's case with better lighting.

Quick recap for anyone who spent Labor Day weekend offline, which, good for you.

August 31. The Federal Trade Commission and 22 state attorneys general file suit in federal court in Seattle. 181 pages. Six years of investigation. Roughly 1.5 million pages of documents. The allegation: starting in 2018, Amazon quietly rigged the price of its search ads, and in the government's words "likely illegally extracted over 20 billion dollars" from the businesses buying them. California's attorney general puts the undisclosed surcharge at more than $20 billion. The formats named are Sponsored Products, Sponsored Brands and Display Ads tied to shopper searches on Amazon's store and app.

Amazon called the suit "misguided." It said it "strongly disagrees."

Then it did something companies almost never do in week one of a federal case.

It described the machine.

Here is Amazon, in its own response, explaining how its auction actually charges you:

"When the winning advertiser's bid exceeds both the hard and soft reserve, they pay the soft reserve, which is less than they were willing to pay."

And:

"When the winning advertiser's bid exceeds the hard reserve but doesn't meet the soft reserve, we still grant the placement to that advertiser and they pay their bid."

They pay their bid.

That second sentence has a name. It is called a first-price auction. Not a modified second-price auction. Not a relevance-weighted marketplace. You bid two dollars, you pay two dollars.

Now put the FTC's number next to it. According to the complaint, Sponsored Products advertisers paid their full bid roughly 30% to 40% of the time in 2021. About 70% in 2022. About 80% in 2024.

So on Amazon's own description of its own rules, the "you pay your bid" branch fired four times out of five. Which means for most winners, the soft reserve, Amazon's own estimate of what the slot was worth, sat above what the advertiser was willing to pay.

A reserve that sits above 80% of winning bids is not a floor under the market. It is the price list.

Amazon's best line, and it will be on every slide in every deposition:

"In no scenario does an advertiser pay more than their bid."

True.

Also true of every first-price auction ever run, including a county fair cattle sale.

Nobody accused Amazon of charging more than the bid. The promise Amazon made, according to the FTC, was never "you won't pay over your max." The promise was a penny more than the next guy. Ads rank on bid and relevance, the winner pays what it takes to beat the runner-up. That is the second-price deal. That is the deal the bamboo cutting board seller thought he had.

An agency owner at eComCatalyst walked through it on YouTube this week with his own accounts. Bid $2. See an average CPC of $1.76. Assume a competitor must have been sitting around $1.74 or $1.75. "And what we're learning is no, that's not exactly the way it was working."

His advice to every Amazon seller now: treat your bid like a price Amazon may charge you in full, "because according to their own explanation, they can and they will."

That is not a plaintiff's lawyer. That is a customer.

Here is the language the FTC says came out of Amazon's own internal materials. These are allegations, and the documents haven't been shown to the public in context. They are also brutal.

Amazon's pricing was described as "a proxy 2nd price that we calculate."

One Amazon scientist allegedly called the mechanism "an invented auction participant."

Another document said the pricing had a "surcharge hidden in it."

And, per the FTC, somebody described it as a "clever non-transparent way to charge first price."

The complaint says Amazon knew disclosure would hurt. Employees allegedly warned it could cause "irrevocable damage to advertiser trust" and a "downward spiral" of advertisers cutting bids and spend. The FTC says Amazon tested how high the surcharge could go without advertisers noticing, pushed it harder on Prime Day and Black Friday, and gave misleading answers when advertisers asked directly whether the auction format had changed.

Amazon's response to the emails:

"A stray email is not indicative of a team's intent or even collective viewpoint."

And: "People at times use email as a brainstorming and suggestive medium."

Sure. A brainstorm does not move a number from 35% to 80% over three years. Brainstorms don't ship. Brainstorms don't get ramped for Black Friday.

And my favorite pairing in the entire record. Amazon, defending itself:

"Reserves are determined in real time and aren't predictable in advance by anyone, including Amazon or the advertiser."

Amazon, per the FTC, internally:

"A proxy 2nd price that we calculate."

Nobody can predict the number. Including the company that calculates it.

Yes. Just not where the headlines put it.

The $20 billion is an opening bid. It hasn't been proven. It will be fought over for years.

The story is the word "auction."

Sarah Caputo told AdExchanger that Amazon spent years telling advertisers they'd pay a penny over the next-highest bidder, "and the FTC says that stopped being true. That's not a technicality." An entire generation of bidding strategy, she said, was built on that one assumption, "and most advertisers never had a way to check it themselves."

Her kicker is the one to forward: "It's what happens when an industry lets 'auction' become a marketing word instead of a defined term."

A former media buyer, unnamed, to the same outlet: "We wouldn't accept this behavior on Wall Street from a trade clearing process. We shouldn't accept it on Madison Ave, either."

And another buyer on why the industry went quiet: you'd need your win logs and the losing bids for every auction. "That's basically impossible to get, and the only ones that would have that are Amazon and now the FTC via subpoena."

Then James Deaker, AKA the self-proclaimed “the Yield Doctor,” who was at Yahoo when its exchange switched to first price in 2019, asked the only question that matters:

"If Amazon wanted the economics of a first price auction, why didn't it simply tell advertisers that it was running one?"

The entire open web moved to first price by the end of 2019. Out loud. With announcements. Google Ad Manager moved. Magnite, OpenX, Index Exchange moved earlier. Buyers adjusted.

Amazon, the FTC alleges, went the other direction. First-price economics. Second-price label.

Why keep the label?

Amazon's own documents already answered that. The label was worth money.

Amazon's rebuttal is stuffed with statistics. Every one of them is real. Almost every one of them answers a question the FTC didn't ask.

Start with relevance.

"In 2024, approximately 92% of selected Sponsored Products ads were not the highest bid, often by a wide margin."

"The mean winning advertiser's bid is typically about the 128th bid by amount."

Great. Genuinely. That means Amazon's auction weighs relevance heavily, and the highest bidder doesn't just buy the slot. Nobody is suing over that.

That is allocation. The lawsuit is about pricing.

Who wins and what the winner pays are two different questions. Amazon keeps answering "who won." The FTC asked "what did they pay." A relevance model that picks the 128th-highest bidder can still charge that bidder a surcharge Amazon invented. The two facts coexist without any friction at all.

Next, the savings.

"Even accepting the FTC's flawed premise that advertisers do not adjust bids, we estimate they saved over $8 billion from 2021 to 2025 as a result of Amazon prioritizing ad relevancy over selecting ads on bid price alone."

Saved compared to what? Compared to an auction where the highest bid always wins. An auction nobody proposed. An auction nobody accused Amazon of running. The 2026 projections work the same way: "at least 58% higher sales" and "at least 46% better return on ad spend" versus that same imaginary highest-bid machine.

Amazon is taking credit for not robbing you in a way nobody said it robbed you.

Now the two statistics that matter, both straight from Amazon.

"Average winning bids fell 50% from 2019 to 2025 on Sponsored Products search ads."

"From 2019 through 2024, the average cost-per-click for Amazon's Sponsored Products search ads remained flat adjusted for inflation."

Amazon offers these separately, as good news. Put them in the same sentence.

If the average winning bid dropped by half and the price paid per click held flat in real terms, then the share of the bid that advertisers actually paid roughly doubled. If that bid figure is in nominal dollars, which Amazon doesn't specify, the real ratio climbed by more than double. The windows are off by a year, 2019 through 2024 for CPC and through 2025 for bids, so treat it as directional. Directionally, it's a cliff.

The gap between what you said you'd pay and what you paid collapsed.

That is exactly what happens when a second-price auction starts behaving like a first-price auction and buyers slowly figure it out. You pay more of your bid. So you bid less. The bids fall. The price you pay doesn't.

Which lines up neatly with the FTC's escalation curve: 30% to 40% paying their full bid in 2021, 80% by 2024.

Amazon's own numbers don't refute the complaint. They chart it.

Deaker made a point on video this week that almost nobody else has picked up, and it is the key to the whole damages fight.

When exchanges first flipped to first price in 2017 and 2018, the early movers made extra money. Why? "The bidders hadn't really adapted to needing to shade bids down." Buyers kept bidding as though they were in a second-price auction, paid inflated prices, and the exchange pocketed the difference until the market caught up.

Now read Amazon's central defense with that in mind:

"Advertisers adjust bids based on real-world performance, not descriptions of auction mechanics."

And: "Advertisers optimize their campaigns based on actual auction outcomes, what they pay, what they win, and the performance they see."

Fine. Let's accept it. Advertisers eventually adjusted. How? By watching costs rise, returns thin and bids stop working the way they used to, and then cutting.

That isn't a defense. That's a description of how the surcharge was discovered: by invoice.

The overcharge doesn't live in the equilibrium, once every seller has learned to shade. It lives in the years before they learned. If you change the rules and don't tell anyone, the market eventually reprices, and you keep everything you collected on the way there. The FTC doesn't have to prove advertisers were still fooled in 2025. It has to show they were fooled while the dial was turning.

Deaker's own framing: the open web cleaned up its auctions around 2019 with "radical first price transparency," while Amazon was allegedly "secretly engineering a pseudo first-price auction behind a second price marketing promise." Same economics. Opposite disclosure. That difference is the case.

The eComCatalyst owner says ad spend as a share of total sales, TACoS in seller-speak, used to run 7% to 8% and you'd be fine. Then 10%. Now he's telling clients 12% to 15%. "It's crept up higher and it's crept up higher."

Amazon says CPC was flat.

Both can be true, and fairness demands saying so. TACoS can climb because Amazon added more ad slots, pushed ads into more placements, or crowded out organic listings, and he complains about exactly that, ads landing in placements that don't convert. Rising TACoS is not proof of a soft reserve. Anybody publishing it as proof is doing Amazon's lawyers a favor.

And there's a real counterweight from the buy side. Luke Stillman of Madison and Wall told Business Insider that "advertisers decide where to spend based on whether the advertising works and the return they're getting from their campaigns. That is more true in commerce media than in any other channel."

Translation: budgets aren't leaving. Zak Stambor of EMARKETER put it more bluntly: "Advertisers face a tough challenge because Amazon is incredibly hard to walk away from."

That is the part Amazon is counting on. Not that it's innocent. That you have nowhere else to go.

Amazon's clearest disclosure argument: "The Ad Console Campaign Builder... has clearly stated since 2018 that a bid represents the maximum an advertiser could be charged."

A maximum is true in a first-price auction and in a second-price auction. It tells you the ceiling. It tells you nothing about how the price under that ceiling gets formed. It is a disclosure that discloses nothing about the thing in dispute.

Then the training materials. Amazon says the FTC "cherry-picked" a few low-reach videos and courses with "older or simplified examples." Three courses had 1,849 enrollments and 779 completions over their entire lifetimes. One video had 928 viewers in two and a half years, "approximately 0.09% of U.S. Sponsored Ads advertisers."

Do that division. 928 is 0.09% of roughly a million U.S. advertisers. Thank you, Amazon, for confirming the size of the class.

Three problems with the low-reach defense.

One. People who take Amazon Ads courses aren't a random sample of sellers. They skew toward the agency staff and tool-vendor employees who set bids for thousands of accounts. A small enrollment number can sit upstream of a very large amount of spend.

Two. The FTC isn't only citing videos. It alleges sales presentations, and it alleges misleading answers to advertisers who asked directly whether the auction had changed. There is no low-reach defense for the answer you gave the person who asked.

Three. Amazon says of those materials: "When we discovered them, we either removed or updated them." Amazon discovered Amazon's training. And it says that after the FTC raised concerns, "we updated our Amazon Ads Help content to explicitly explain the use of reserve prices." That's the help page we flagged on August 31, the one updated this spring.

You don't rewrite the manual in 2026 to explain something everybody already understood.

Amazon's lawyers already know exactly where the FTC's case is soft. You should too. Below the line: the arguments Amazon will actually win, the one internal warning that could sink all of them, and why the states, not the FTC, are holding the checkbook. Then the part other retail media networks hope nobody reads: who else is sitting on the same second-price label Amazon just got sued over, the AI ad marketplaces that defaulted straight into it, and the Google case already warming up in the bullpen.

After that comes the part your agency hopes you never read: a six-step checklist for building your own version of the FTC's 80% number from data you probably already have, what to preserve before you need it, and the contract language every RMN should be made to sign. Amazon says advertisers learn from real-world performance. Upgrade to ADOTAT+ and learn from something cheaper than your invoice.