The most consequential algorithm on Amazon this year is not Alexa for Shopping. It is the one that decided what you paid for a click, and on August 31 the FTC described how it allegedly worked.
If the complaint is right, the number you type into the bid field has not been a ceiling you rarely touch. On roughly four out of five Sponsored Products clicks in 2024, it was the price. That changes how you should think about every bid multiplier, every dynamic bidding setting, and every event-based rule you are about to switch on for Prime Big Deal Days and Black Friday. Not because Amazon has been caught, which a court will decide, but because the mechanism it describes is a better explanation for what your ACOS has done since 2021 than anything in your agency's monthly deck.
What happened
On August 31, 2026, the FTC and attorneys general from 22 states filed a 181-page complaint in the Western District of Washington (FTC press release, covered by TechCrunch the same day) alleging that Amazon told advertisers Sponsored Products ran a second-price auction, then from 2019 added an undisclosed "soft reserve price" that charged advertisers their own bid instead. Amazon calls the suit "misguided," says the complaint "fundamentally misunderstands how advertisers operate," and says average bids fell 50% between 2019 and 2025 while advertisers saved more than $8 billion.
This is eight days old. I sat on it while the complaint text landed and the practitioner write-ups caught up, because the first-day coverage was about the dollar figure and the dollar figure is the least useful thing in it.
Why most brand owners will read this wrong
The dumb take is "Amazon overcharged me, when do I get my refund." You do not, not this quarter and probably not this decade. This will run for years, Amazon will contest every number, and nothing about a refund changes a decision you make in September.
The other dumb take is the mirror: "it's allegations, Amazon disputes it, nothing to see." Both of Amazon's defenses can be true at the same time as the complaint. Average bids can fall 50% while the share of clicks charged at the full bid rises from 4% to 79%. Those are two different numbers describing two different things, and only one of them is about what you paid relative to what you were told you would pay.
The real signal is the mechanism, and it is worth reading slowly because your bidding strategy was built on the opposite assumption.
A second-price auction means you bid what a click is worth to you and pay one cent above whoever came second. The whole reason that design exists is so advertisers bid honestly instead of shading down. Amazon documented that this was how Sponsored Products worked. The complaint says that starting in 2019, after the auction had already found a winner and a natural second price, a second calculation ran. Internal documents quoted in the filing call it an "invented auction participant," a synthetic bid representing what Amazon thought the slot was worth, capped only by your winning bid. One executive is quoted describing the result as a "proxy 2nd price that we calculate." Another document calls it a "clever non-transparent way to charge first price."
The numbers the FTC puts on the outcome, for Sponsored Products specifically: the share of clicks where the advertiser paid their own bid was about 4% in late 2020, 30 to 40% in 2021, roughly 70% in 2022, and 79.1% in 2024.
Read that as an operator, not as a plaintiff. If the share of clicks priced at your bid went from one in twenty-five to four in five over three years, then your bid stopped being an upper bound and became your price, and it did so gradually enough that no single month looked like a repricing. It looked like drift. It looked like the market getting more competitive. It got explained, in a thousand agency calls including some I was on, as "CPCs are up across the category."
What changes for someone running $200K a month on Amazon
Two weeks ago I derived from Amazon's Q2 report that advertising revenue grew about 8% per paid unit year over year while the fee side barely moved. I said then that the ad half of Amazon's take is priced by an auction nobody announces, and that it arrives in your ACOS looking like a report card on your own execution. The complaint hands that argument a specific mechanism. I am not going to claim it proves the derivation. I am going to say the shapes match.
Here is what it changes, concretely.
Your bid is a price, not a ceiling. On a $30K a month account, the standard move for four years has been to bid above your target CPC on important terms, trusting the second price to bring the actual cost down toward the market. Under the alleged mechanism, on most clicks there is no market below you. There is your bid. A brand bidding $2.40 on a term where it was willing to pay $1.90 has, on four out of five clicks, simply paid $2.40. That is a 26% overpayment built into the strategy, not into the auction.
Every multiplier is now a direct CPC increase. Top-of-search placement modifiers, event-based bid rules, dynamic bidding up-and-down. All of them raise your bid on the theory that a higher bid buys position and the second price protects the cost. If the second price is not there, a 50% peak multiplier is a 50% CPC increase on most of the clicks it wins. Amazon's up-and-down setting can raise your bid up to 100% at top of search. Under a first-price reality, that is not a bid adjustment. It is a price you agreed to in advance without seeing it.
The ACOS-up, share-flat quarter has a candidate explanation. I wrote about impression share as the only Q4 metric that describes the market rather than your account. A rising ACOS with flat impression share means you paid more for the same ground. The complaint describes a mechanism that produces exactly that pattern, account-wide, for years, with no change in competitive intensity required. An account manager quoted in the filing relayed a grocery client saying CPCs were up more than 90% and killing return on ad spend. Nobody in that conversation had the mechanism either.
Q4 amplifies it. Peak is when every brand raises bids simultaneously and when you are most likely to have automated rules doing it for you. If the price you pay is your bid, the week your rules fire is the week the gap between "what I'd pay" and "what I typed" costs the most. On $60K of November spend, a 20% gap between willingness to pay and bid is $12,000 that no report will label as anything other than "peak CPCs."
What I would do this week
Measure your own bid-to-CPC ratio. Pull the last 30 days of search term or targeting reports for your top 20 revenue terms. Divide average CPC by your bid on each. If most of them sit above 0.9, the second-price cushion is not there for you, whatever a court eventually decides. Save the file with the date in the name. This is the one thing that becomes harder to reconstruct later, because September is the last month of normal traffic before your bids start moving.
Rewrite peak bid rules as price statements. Not "raise bids 40% during the event." Instead: "we will pay up to $X for a click on this term at this conversion rate." Then set the bid to that number, because that is what you will pay. If the resulting bid does not buy the position you wanted, that is information about what the position costs, not a reason to raise the bid.
Re-decide dynamic up-and-down before October, campaign by campaign. On any campaign where your ratio is already above 0.9, up-and-down is handing Amazon the authority to double your price at top of search. Fixed bids or down-only with your own placement modifiers give you the same reach with a price you set.
Ask your agency one question. "What share of our clicks last month were charged at or near our bid?" A shop that has pulled the data answers with a number. A shop that answers with a paragraph about auction dynamics has not looked, and that is worth knowing before they run your peak.
Stop grading Q4 on ACOS alone. I have said this three times this year and the complaint is the strongest argument yet. In the quarter where every input moves, an efficiency ratio tells you what the market did and invites you to take it personally. Add ad cost per unit sold and impression share, and read all three together.
What I would ignore
The dollar figure. "Tens of billions" per the FTC, "$20 billion plus" in some write-ups. It is a headline number about seven years of a $68 billion a year business. It changes nothing about your bid on Thursday.
The refund discourse and the class-action solicitations that will arrive in your inbox by October. Anyone selling you a recovery service today is selling a position in a queue that does not exist yet.
Amazon's "inflation-adjusted CPC was flat" defense as a reason to relax. A flat average CPC is entirely consistent with your bids falling and your paid-to-bid ratio rising. That is the point.
The urge to move budget off Amazon in protest. Your customers are still on Amazon. The lesson is not to leave the auction, it is to stop bidding as if the auction were something it may not have been.
Anyone who tells you the case changes what a bullet point should say. It doesn't. This is a pricing story, and the only creative implication is the one that was always true: CTR and CVR decide how much a click is worth to you, and that number, not a multiplier, should be what you type in the box.
Five or six times this year I have written about something moving underneath an operator who did nothing wrong. A model swapped, a price expired, a title got rewritten, a statute took effect in June with no changelog. This one is older than all of them. If the complaint is right, the auction under your account changed in 2019, and for seven years the only evidence was a number in your own console that everyone, me included, read as the market. Go pull the ratio. It is the one artifact in this story you actually own.