The Product You Wanted Was on Page Four

Based on the research ofKe, Lin, and Lu, "Information Design of Online Platforms," Management Science, 2026

Ke, Lin, and Lu's paper in *Management Science* models something every online shopper has felt and never had a name for: the platform showing you products is not simply trying to match you with the best one. It is running two businesses at once, helping you search, and helping sellers advertise, and those two jobs pull in opposite directions. A platform that always surfaces the single best match kills its own auction, because no seller needs to pay for prominence when the algorithm will find them anyway on the merits. So the optimal information design, in the authors' model, does something colder: it deliberately mixes the matched product with a long tail of unmatched ones, limits how much a consumer can learn from a single search, and lets that engineered uncertainty become the thing sellers bid to escape. The paper's own words for the result are blunt: the design "may be socially inefficient," and "sponsored targeted advertising on retail platforms may introduce match inefficiency." Read plainly, that is an argument that the ad business and the discovery business are in conflict, and platforms have a mathematical reason to let the ad business win a little.

The Chain: Blur, Compete, Extract

The mechanism runs in three moves. First, the platform blurs the signal, instead of ranking purely on fit, it recommends a mix that includes the true best match somewhere inside a wider set of look-alikes, so no single result is obviously "the" answer. Second, sellers now have to compete for the prominence the algorithm withheld, which means bidding against each other for placement rather than simply being found. Third, the platform captures that competitive spending as auction revenue, while the buyer, on average, ends up clicking something a little worse than what a purely honest recommendation would have shown. None of this requires malice or a rogue engineer. It falls out of the platform maximizing its own revenue subject to sellers behaving rationally, which is exactly why it is durable: nobody has to decide to do it badly, the incentive gradient does the work.

This is a sharper claim than the familiar complaint that "sponsored" labels are confusing. The paper is not describing an ad product bolted onto a neutral search engine. It is describing a single design choice that determines both what you see and how much sellers pay, meaning the platform's profit-maximizing search quality is mediocre by construction, not by neglect.

What the Blur Looks Like on Amazon

You do not need the model to see the shape of it; regulators already photographed it. The FTC's September 2023 complaint against Amazon, filed with 17 state attorneys general, alleges the company was "degrading the customer experience by replacing relevant, organic search results with paid advertisements, and deliberately increasing junk ads that worsen search quality," while separately "biasing Amazon's search results to preference Amazon's own products over ones that Amazon knows are of better quality." The same complaint alleges that when Amazon detects a seller pricing lower elsewhere, it "can bury discounting sellers so far down in Amazon's search results that they become effectively invisible", the long tail, weaponized against anyone who might otherwise win on price. Combined, the FTC alleges these tactics let Amazon extract fees running "close to 50% of total revenues" from the sellers who depend on that search page for their livelihood.

The Markup's 2021 investigation supplies the texture. Reporters analyzed Amazon's results for 3,492 popular searches and found that knowing only whether a product was an Amazon house brand or exclusive predicted the top search spot in 71 percent of cases, a better predictor than the product's star rating (55 percent) or its number of reviews (52 percent). Amazon's own cinnamon cereal, four stars and 1,010 ratings, sat in the number-one slot ahead of Cap'n Crunch at five stars and 14,069 ratings. Eighty-seven percent of the Amazon-brand listings placed first were tagged "sponsored" in the page's source code, a status never disclosed to the shopper reading it. And sellers describe the escape hatch exactly the way the model predicts: "You turn off the ads and you lose organic rank within days," Amazon consultant Jason Boyce told The Markup. "It's pay to play." That is stage two of the chain, seller competition for prominence, quoted verbatim by the people paying for it.

Sponsored targeted advertising on retail platforms may introduce match inefficiency.

The Auction Behind the Auction

If burying results is the blur, Amazon's advertising auction is the extraction stage laid bare. In an August 2026 complaint, the FTC and 22 state attorneys general allege that for more than seven years Amazon told advertisers it ran a standard second-price auction, where the winner pays one cent more than the runner-up bid, while secretly inserting what internal documents call an "invented auction participant" and a hidden "soft reserve price" that made advertisers pay their own full bid close to 80 percent of the time by 2024, up from 30 to 40 percent in 2021. The complaint quotes an Amazon employee explaining the point of the trick with unusual candor: the surcharge lets Amazon capture prices "beyond what \[can\] be achieved through advertiser competition," and is "good for Amazon" precisely because "the benefit to Amazon comes at the cost of advertisers." Whatever the legal outcome, the internal language matches the paper's mechanism move for move, a platform quietly designing the terms of competition so that the money sellers spend fighting for prominence exceeds what an honest, transparent auction would ever produce.

Google Already Paid for the Proof of Concept

Amazon is not the instructive case because it is unusual; it is instructive because a second dominant platform ran the identical play in a different market and a regulator already finished the math. In 2017 the European Commission fined Google €2.42 billion (precisely €2,424,495,000) for, in its words, "systematically" giving prominent placement to its own comparison-shopping service while demoting rivals through its general search algorithm. The Commission's own evidence, drawn from 5.2 terabytes of real search data, found that the most highly ranked rival shopping service appeared, on average, only on page four of Google's results, a demotion so severe that traffic to rival services fell 85 percent in the UK and 92 percent in Germany, while Google's own comparison service gained traffic up to 45-fold in the same markets. The Commission also quantified exactly why placement is the whole game: the top organic result captures roughly 35 percent of all clicks, and the first result on page two gets about 1 percent. Bury a competitor from page one to page four and you have not demoted it. You have deleted it, and channeled the customers who would have found it toward the paying product instead, blur, compete, extract, at continental scale, with a court-upheld price tag attached.

What the Rule Actually Predicts

None of this requires assuming Amazon or Google set out to harm consumers; the paper's contribution is showing that the harm is the profit-maximizing outcome of an ordinary optimization problem, which is a worse finding for the industry than intent would be, because it means the pressure toward blur does not go away when better people run the algorithm. Any platform that both recommends and sells prominence is solving the same equation, whether it is a grocery delivery app burying the cheapest staple beneath three sponsored substitutes or a freelance marketplace nudging the best-reviewed contractor two rows down. The paper's mechanism chain gives operators, regulators, and shoppers the same diagnostic question: when a platform's revenue depends on sellers competing for visibility, ask what the algorithm would need to hide from you for that competition to exist at all. The FTC's complaints and the EU's fine suggest the honest answer, in at least two of the world's largest platforms, was your best match.