The price to beat in Amazon’s advertising auctions may not have come from another advertiser at all, according to a lawsuit filed Aug. 31 by the Federal Trade Commission (FTC). In the complaint, the FTC alleges that Amazon quietly set the price itself, promising one pricing rule and using another.
That allegation could support a conventional deception claim. If Amazon promised advertisers a genuine generalized second-price auction while deliberately concealing a material departure from that model, the FTC may have a viable case.
But the complaint tries to make a broader leap, arguing that Amazon’s pricing practices necessarily harmed retail consumers. The use of reserve prices, relevance scores, or dynamic pricing does not automatically make an auction unfair. Nor does it necessarily harm the consumers who buy the advertised products.
That is the complaint’s weaker theory. Based on the public record, the FTC has asserted retail-consumer harm with confidence but has yet to prove it.
What the FTC Alleges—and Amazon Disputes
The FTC and the attorneys general of 22 states filed the case in the Western District of Washington. The FTC’s press release describes a seven-year scheme that affected roughly 1.2 million advertisers, including more than 500,000 small and medium-sized businesses.
Amazon’s three principal Sponsored Ads products auction placements tied to product searches and product pages. According to the agency, Amazon told advertisers that winners would pay only the minimum needed to secure a placement—often one cent more than the next-highest-ranked bid. Instead, Amazon allegedly imposed undisclosed “soft reserve” prices or surcharges. The complaint estimates that the practice extracted more than $20 billion from advertisers.
The complaint tells a powerful story. It describes an “invented auction participant,” a “proxy 2nd price,” internal concerns about advertiser expectations, and alleged efforts to keep price increases below detection thresholds. It also alleges that the share of Sponsored Products auctions in which advertisers paid their full bids reached about 80% by 2024.
If proved, these allegations would amount to more than a complaint about hard bargaining. The FTC’s case would be that Amazon promised advertisers one pricing mechanism while secretly operating another.
Amazon offers a different account. The company says its system evolved from rankings driven heavily by bids to machine-learning models that give greater weight to relevance and predicted shopper engagement. It describes soft reserves as real-time minimum prices that reflect the value of individual placements, much as a retailer might value an endcap differently from ordinary shelf space.
Amazon also says its campaign console has long informed advertisers that a bid represents the maximum they may pay. According to the company, the highest bidder does not win roughly 92% of placed Sponsored Products ads, average winning bids fell by 50% from 2019 to 2025, and both conversion rates and return on advertising spending improved. These claims are not adjudicated facts, but they identify the central economic issues the complaint must confront.
The legal questions are narrower than whether Amazon is large, profitable, or vertically integrated. Did the company materially misrepresent its price-setting rules to reasonable advertisers? Did its practices cause legally recognized consumer harm that advertisers could not reasonably avoid and that countervailing benefits did not outweigh? Those inquiries should not collapse into a general judgment that Amazon earns too much from advertising.
How Advertising Auctions Work
Advertising platforms use auctions to allocate scarce, varied opportunities. A search page has room for only a limited number of ads, and some positions attract more attention than others. A placement’s value also depends on the query, nearby products, the shopper, the time of day, and the likelihood of a click or purchase. A pricing mechanism that treats every opportunity alike may be simple, but simplicity does not guarantee efficiency.
In a single-item second-price auction, the highest bidder wins and pays slightly more than the second-highest bid. Under the classic Vickrey model, each participant’s best strategy is to bid what the item is worth to them.
Digital-advertising auctions are more complicated. As Benjamin Edelman, Michael Ostrovsky, and Michael Schwarz explain, a generalized second-price auction with several advertising slots is not merely a series of Vickrey auctions. Because the slots produce different click-through rates, bidding one’s true value is generally not the dominant strategy. Advertisers may lower their bids, experiment with different amounts, or use automated systems to maximize expected returns.
That distinction matters. The FTC repeatedly treats “second price” as though it describes a complete and uniquely determined mechanism. But a relevance-adjusted auction requires rules governing eligibility, quality, expected click-through rates, conversion probabilities, placement, reserve prices, and payment. A platform may use a second-price-like payment rule while considering nonprice information when allocating placements. It may also decline to sell a placement below a specified floor. A reserve price is not inherently a fake rival. It is a seller-side condition on the transaction.
The harder question is whether Amazon disclosed that condition. A hidden reserve may alter advertisers’ bidding incentives if they believe the second-highest competitor determines the price and therefore bid close to what a placement is worth to them. A disclosed reserve may simply prompt an advertiser to bid only when the placement is worth at least that amount.
The significance of Amazon’s alleged conduct therefore depends on the counterfactual. What did advertisers reasonably believe? What did Amazon’s system actually calculate? How did the difference affect bids, campaign spending, and realized returns? Calling the reserve a “shill bid” is evocative, but the legal and economic analysis should turn on those mechanisms and effects, not internal labels.
The FTC’s Deception Case
Section 5(a) of the FTC Act prohibits unfair or deceptive acts or practices (UDAP) in commerce. Under the agency’s longstanding Policy Statement on Deception, a practice is deceptive if it is likely to mislead a reasonable consumer about something material—meaning something likely to affect the consumer’s conduct regarding a product or service. When the audience is specialized, the standard reflects a reasonable member of that audience, not an unsophisticated retail shopper.
Applied to Amazon’s advertising customers, that framework gives the FTC a potentially strong argument. “Second-price auction” and statements that the winner pays the second-place bid are not mere puffery. They describe a pricing rule that can affect bidding strategies, budget allocation, expected customer-acquisition costs, and the value advertisers assign to campaign data.
The complaint alleges that Amazon repeated these claims on websites, in training materials, and during sales presentations even after changing its auction system. It also claims that Amazon gave evasive or false answers when advertisers asked whether the format had changed. A factfinder could view that pattern as a material misrepresentation, not a harmless simplification.
Amazon’s response points to four principal defenses. First, the campaign console allegedly told advertisers that a bid represented the maximum price they could pay. Second, Amazon says its system was not a simple highest-bid auction because lower bids could win based on relevance. Third, the company describes the materials cited by the FTC as outdated educational content with limited reach that it later removed or revised. Fourth, Amazon argues that sophisticated advertisers and automated bidding tools respond to actual clicks, conversions, costs, and returns—not simplified descriptions of auction theory.
Those defenses do not dispose of the case. Telling advertisers that a bid sets the ceiling does not necessarily disclose that Amazon may substitute an internally generated floor for the competitive price. Nor does advertiser sophistication give a seller license to make a specific false claim about pricing.
But the defenses sharpen the inquiry. The question is not whether some Amazon document used the phrase “second price.” It is whether the overall impression created by Amazon’s communications, viewed in context and at the time of each transaction, conveyed a materially false pricing rule to reasonable advertisers.
Even if the evidence supports a deception finding, monetary relief requires more. Any estimate of advertiser or consumer injury must show that the alleged misunderstanding changed behavior and caused a measurable loss, not merely confusion over terminology.
The FTC’s Unfairness Case
The FTC’s unfairness theory faces a different statutory test. Section 45(n) bars the agency from declaring a practice unfair unless it causes or is likely to cause substantial consumer injury that consumers cannot reasonably avoid and that countervailing benefits to consumers or competition do not outweigh. The FTC Policy Statement on Unfairness makes consumer injury the central concern. It also requires the agency to consider offsetting benefits and the costs of intervention, including reduced incentives for innovation and capital formation.
The complaint offers little evidence of harm to retail consumers. Its factual narrative focuses overwhelmingly on advertisers. Consumers enter the story mainly as buyers of products from Amazon merchants. The FTC reasons that merchants likely pass higher advertising costs along to them, especially when selling low-margin necessities.
But the complaint’s ultimate allegation of consumer injury is conclusory. It identifies no retail-price study estimating pass-through, no comparison of advertised and unadvertised product prices, no evidence of reduced output, and no net-welfare analysis that accounts for the relevance improvements Amazon says accompanied its pricing changes.
Amazon highlights precisely this gap. The company says consumers appear only a handful of times in the complaint’s more than 150 pages, the FTC cites no data showing higher consumer prices, and the proposed monetary relief would go to advertisers rather than shoppers. Amazon also reports that inflation-adjusted cost per click (CPC) remained flat while conversion rates and advertiser performance improved substantially.
Those figures require scrutiny. But the FTC cannot prove a consumer-protection case by assuming that every extra dollar an advertiser pays becomes an extra dollar on a shopper’s bill.
Pass-through is an empirical question. An advertiser may absorb higher costs through lower margins; cut its advertising; move spending to Google, Meta, Walmart, direct sales, or other channels; change its product mix; or raise prices only where demand allows. Better-targeted advertising may also reduce the effective cost of acquiring customers even if the nominal CPC rises. More relevant ads may help consumers discover new products, intensify competition among sellers, or reduce search costs.
A practice may therefore transfer surplus from advertisers to Amazon while improving, diminishing, or leaving consumer welfare unchanged. The FTC cannot treat “passed on” as its conclusion when pass-through is the disputed causal link.
That does not make advertiser injury irrelevant. Advertisers are themselves consumers of Amazon’s advertising services, and deceptive pricing claims may harm them directly. But the unfairness theory still must satisfy Section 45(n). A court should demand evidence of substantial net injury, consumers’ realistic ability to avoid that injury, and the benefits of Amazon’s dynamic allocation system.
If the FTC proves deception, a separate unfairness balancing test should not excuse it. But if the agency claims that a commercially sophisticated pricing system is inherently unfair, the statutory cost-benefit test is indispensable.
Choosing the Right Counterfactual
The case also illustrates a recurring problem in platform regulation: treating the regulator’s preferred counterfactual as the natural baseline. The FTC compares Amazon’s charges with the prices a genuine generalized second-price auction would have produced. That may be the right counterfactual for a deception claim if Amazon promised precisely that mechanism. It does not necessarily tell us whether Amazon’s evolving pricing and allocation system worked efficiently.
Harold Demsetz’s critique of the “Nirvana” approach applies directly. Demsetz argued that analysts should compare institutions as they can actually operate, not measure an imperfect reality against an idealized alternative free of costs, frictions, and unintended consequences. Geoffrey Manne has applied the same point to digital platforms. A regulator can imagine a world in which a platform maintains all its investments and services while accepting rules that reduce their returns. But that world is not a serious alternative unless the analysis considers how the platform would respond.
Amazon’s relevance model may be flawed, inefficient, or poorly disclosed. It may also address a real allocation problem. If bids alone determine rankings, a high-paying but irrelevant ad can displace a lower-paying ad that shoppers would find more useful. A relevance model may allow a lower bid to deliver more value to both the shopper and the advertiser. A soft reserve may then represent an imperfect attempt to keep scarce premium placements from becoming systematically underpriced as the ranking algorithm changes.
The relevant comparison is not an abstract choice between “auction” and “no surcharge.” It is Amazon’s actual model compared with feasible alternatives, such as transparent reserves, higher upfront fees, a first-price auction, less investment in relevance, or a different mix of advertising inventory and retail prices.
Friedrich Hayek offers another caution. Amazon’s engineers and economists observe billions of auctions involving changing queries, products, and shoppers. Their models draw on dispersed, context-specific information that no public official can reconstruct from a complaint and a handful of documents. That knowledge gap should inspire humility about prescribing a single permissible auction formula.
It does not give Amazon license to mislead. A market-process perspective supports accurate disclosure and accountability for deliberate deception. It also resists the assumption that regulators can safely dictate how a platform prices every distinct placement.
Ronald Coase likewise urged decision-makers to compare institutional arrangements and their transaction costs. The FTC identifies costs to advertisers and possible downstream effects on shoppers. It must also consider the costs of intervention, including new compliance systems, less experimentation, poorer allocation, higher fixed fees, lower-quality ads, and the risk that platforms will stop serving smaller advertisers.
The central policy question is not simply whom to label the wrongdoer. It is which feasible arrangement produces the greatest expected value after accounting for all relevant costs and benefits.
The Risks of an Overbroad Remedy
Single-firm cases are especially prone to false-positive errors because the same conduct may look like exploitation or innovation depending on facts that outsiders struggle to observe. The error-cost principle counsels caution when intervention may suppress productive conduct that market competition would otherwise test and correct. Frank Easterbrook’s classic analysis and Jonathan Barnett’s discussion of false positives both emphasize that speculative intervention can produce real harms.
The risk here extends beyond an order requiring clearer disclosures. The litigation could evolve into a command that Amazon use a fixed generalized second-price formula, price every placement according to the next advertiser’s bid, or seek regulatory approval before setting reserve prices. Such a remedy could freeze a technology designed to evolve.
Amazon might respond by giving relevance less weight, raising subscription or seller fees, reducing free tools, restricting access for small advertisers, or adopting an openly first-price model in which strategic bid shading becomes central. None of those effects is certain. That uncertainty is precisely what makes a broad remedy hazardous.
Dynamic pricing also serves legitimate economic purposes. It can direct limited inventory toward its highest-value uses, reflect peak demand, and help a platform recover investments in search, logistics, fraud prevention, measurement, and machine learning. A reserve may act as a price floor, a quality screen, or a way to avoid selling a premium placement for less than its opportunity cost.
Those functions may coexist with deception if the platform lies about them. The remedy should target the lie, not treat the underlying pricing flexibility as inherently suspect.
The FTC’s burden therefore breaks into three questions. What did Amazon promise? What did it actually do in each transaction, and how did that differ from the promise? What injury resulted after accounting for the value and performance of the advertising delivered? The first two questions are primarily factual and legal. The third is economic.
A $20 billion gap between an idealized generalized second-price outcome and Amazon’s actual revenue does not automatically equal $20 billion in social loss. Some of that amount may reflect better allocation, avoided costs, valuable impressions, or services that advertisers continued to use voluntarily. After accounting for those benefits, net social welfare may even exceed what it would have been under the supposedly ideal pricing model.
A Remedy That Fits the Harm
The case should not be dismissed merely because the buyers are businesses or because Amazon faces competitors. A material, undisclosed departure from a promised pricing rule can undermine voluntary exchange, especially when the seller controls the data needed to verify the auction. If the evidence shows that Amazon knowingly misrepresented its auction as second price to induce higher bids—and that remains a very big if—targeted disclosure and restitution for the resulting overcharges may be justified. The FTC’s state co-plaintiffs may also seek restitution, even where the agency itself cannot.
Any remedy should remain proportionate. Amazon should have to describe its ranking and pricing system accurately through the interfaces and sales channels advertisers actually use, preserve enough records to permit an audit, and correct materially misleading historical materials.
Monetary relief should rest on a transaction-level counterfactual that separates any overcharge caused by deception from the value created through better relevance and performance. The court should resist requiring a single auction design, banning every undisclosed reserve regardless of context, or treating all of Amazon’s advertising revenue as ill-gotten.
This approach would enforce the law against deception without turning consumer protection into price regulation. It would also leave Amazon room to find better ways to match products, shoppers, and advertisers. The law should insist that market participants know the essential rules of the game. It should not assume that public officials know how best to play it.
Truth Without Price Regulation
The FTC’s lawsuit against Amazon may establish an important principle: sophisticated auction systems must still obey ordinary rules against material deception. But the public record does not yet support the complaint’s broader claims that dynamic reserve pricing necessarily harmed retail consumers or that regulators can safely supervise the design of a complex advertising marketplace.
The outcome should turn on evidence of what Amazon promised, what it concealed, and what economic harm followed. A market-oriented law & economics analysis should demand truthful rules, empirical evidence of consumer harm, and relief tailored to that harm. It should also take seriously the cost of regulatory mistakes that micromanage innovation.
The FTC may have a deception case. It has not yet proved a pricing case.
