When a shopper searches for a hotel, a new phone or a pair of shoes, they may assume the unpaid results are simply the best available matches after any advertisements have been marked. An economics study suggests the incentives can be more complicated. Under the conditions modelled by its authors, a search platform can increase revenue by making ordinary listings less informative while allocating valuable matches to sponsored search results. That is a warning about how a marketplace may be designed, not evidence that every popular website behaves this way.
The research appeared in The Economic Journal in January 2026. Its authors examine how a search platform, product sellers and customers interact when sponsored positions are sold. They do not claim to have audited every live result on Amazon, Google or a hotel-booking site. The work is a mathematical model exploring what a platform might find profitable under particular assumptions. Those limitations matter because the headline finding is provocative enough to be misread as an allegation about individual companies.
What sponsored search results change
A search page looks like a straightforward collection of options, but the platform showing it may be paid in several ways. It can sell a prominent placement to a business, receive a commission when a customer buys something, or earn from related activity. A seller wants to reach a buyer likely to choose its product. The shopper wants a useful choice at an acceptable price, ideally without wasting time opening a long succession of unsuitable listings.
Those interests overlap but are not identical. A platform might know a great deal about which product suits a particular customer, perhaps from a previous search or purchase. If it uses that information to place a strong match in an advertisement, the paid position becomes more valuable to sellers. If the remaining results are unusually easy to compare, however, the shopper may quickly move away from the paid position. The way unpaid information is presented can therefore influence the market for advertising.
Researchers call attention to a strategy in which free listings are made less informative. This does not necessarily mean falsifying a price or removing a product. It can involve a ranking structure that reveals less about relative quality as a person moves through unpaid choices. The exact mechanism depends on the model. What matters is the gap between a platform’s knowledge of the choices and the information it elects to show the shopper at a glance.
A Cornell account of the study explains the counterintuitive result in ordinary terms: a platform may benefit when unpaid results offer less guidance, because customers then have stronger reasons to consider the sponsored position or continue browsing. The model also allows for situations in which a prominent sponsored match helps a customer. Paid placement is not automatically bad, and the welfare consequences depend on the details.
Why a useful advertisement is still an advertisement
It is tempting to divide search results into good unpaid recommendations and bad paid ones. Reality is more complicated. A sponsored result might happen to be an excellent match for the shopper. A free result might be poor, outdated or less relevant. The problem is that the criteria for choosing the paid position can include the money flowing to the platform, while the user may be thinking mainly about product quality or price.
Imagine searching for accommodation near a railway station. Several properties may suit the request. A platform could identify an especially convenient hotel and place it prominently in a sponsored position. That result may genuinely save time. Yet a shopper who sees only the first few entries may not learn whether a similarly suitable alternative was available at a lower total cost. The issue is not the presence of an advertisement alone. It is what the surrounding design helps or prevents the shopper from comparing.
Prices themselves complicate the comparison. A hotel room may carry extra charges. A phone listing may advertise a device price while requiring a contract. Delivery fees and return conditions can matter more than the headline amount. A useful search display helps people see the total proposition; a less informative one makes comparison more laborious. The research addresses incentives for presentation rather than measuring how often any particular undisclosed charge appears.
Why this is a model, not a scandal finding
The economists construct a setting in which many firms compete for shoppers’ attention and sponsored placement. Their results depend on how the platform earns revenue, the information it possesses, what shoppers can learn and how costly it is for them to inspect another result. Changing those assumptions can change the preferred ranking. This is the central limitation of theoretical work: it can reveal a powerful incentive without documenting its frequency in the wild.
Nothing in the model proves that a named retailer has deliberately manipulated a particular user’s results. Nor does it establish that every change in ranking quality is driven by advertising revenue. Search relevance is technically difficult even without commercial considerations. A product may be unavailable, unusually expensive or incorrectly described. Algorithms often juggle numerous objectives, some legitimately connected to customer experience.
However, identifying incentives is useful in its own right. Marketplace design is not neutral merely because it is produced by software. Someone decides which information is considered important, how a result is ranked, where advertisements appear and what a user must click to compare choices. These decisions shape attention and can affect what customers buy. The published study provides a reason to examine those choices rather than treat the screen as an objective list of winners.
Where AI shopping assistants fit
The finding becomes particularly relevant as shopping tools move from lists of links to conversational recommendations. An AI assistant may select a handful of products rather than show a whole page of competitors. That makes the reasoning behind a recommendation harder for the customer to inspect. A useful explanation should distinguish product suitability, availability, price and any commercial relationship affecting placement. Otherwise, a very fluent recommendation can hide ordinary marketing incentives behind the appearance of personalised advice.
LiveAIWire has investigated AI product recommendations that change what shoppers see. That reporting concerns the output of recommendation tools, whereas the economics paper examines why a platform controlling both information and paid space might design its search results in particular ways. The two questions meet at the shopper’s screen, where commercial presentation and machine-generated judgement can be difficult to separate.
A shopper also needs to know whether the assistant searched the whole market or only a selected group of participating merchants. A suggestion may be useful within the available catalogue while omitting competitors outside it. That is not necessarily deceitful if the limitation is clear. It becomes misleading when the tool implies comprehensive comparison without having performed one. Asking what sources were considered is therefore as important as asking which item was recommended.
Practical ways to compare before buying
Consumers cannot inspect a platform’s ranking algorithm, but they can vary where they look. Checking more than one seller, sorting by total price where possible and examining delivery and returns can reveal options that a default list obscures. Sponsored labels indicate commercial placement, not a judgement that the product is unsuitable. The sensible response is neither to avoid every advertisement nor to accept the first convincing result. It is to compare the underlying offer.
Businesses selling online face different incentives. A merchant might feel compelled to pay for visibility because shoppers rarely look beyond the first screen. That can affect competition even when the advertisements are clearly marked. Smaller sellers may lack the resources to win bidding contests, while larger firms may use their budgets to remain visible. The economics model considers platform and consumer incentives, not a comprehensive measurement of outcomes for every type of business.
The LiveAIWire report on Amazon pricing bots looks at another way automated marketplace decisions can affect consumers. Pricing and ranking are distinct decisions, but both show why a product page’s apparent simplicity can conceal interactions among sellers, software and financial incentives.
The real question is whether results deserve trust
Trustworthy product search is not simply a matter of providing many options. It means allowing a person to distinguish a relevant choice from a paid promotion, understand significant costs and see enough information to compare. A platform may legitimately choose to sell advertising while still helping users shop well. The challenge is to make the commercial logic visible enough that relevance does not become an empty word.
This issue will grow as search pages give way to AI agents able to choose, and potentially purchase, on a customer’s behalf. An automated buyer has even less opportunity to notice an omitted competitor than a person scrolling through a page. Setting requirements about budget, preferred merchants, total costs and approval before checkout can make the process more accountable. The system should also explain what it actually compared.
LiveAIWire has covered AI retail assistants helping with purchases, a development that makes the mechanics of recommendation more important, not less. Convenience is useful, but a faster answer is not necessarily a fuller view of the market.
The research’s most useful lesson is subtle. A paid result can be a good result, and a search platform can have an incentive to make the surrounding free information less helpful. Those possibilities should encourage scrutiny, not sweeping accusations. For shoppers, the durable habit is to treat a recommendation as a starting point for comparison, especially when the service providing it can also earn money from the choice.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.