If an AI shops for you, whose interests does the search algorithm actually serve — yours, or the platform's profit?

This explores why a platform's revenue goals can diverge from consumer interests in search and shopping markets, especially as AI agents take over the searching. The corpus has a surprising first answer: the conflict isn't always aimed at consumers. One formal model finds that a profit-maximizing platform prices 'inspection tokens' (the cost of looking at a seller's offer) so that buyers over-inspect. Consumer welfare stays flat, but money moves from sellers to the platform and the market as a whole gets less efficient Does platform pricing of inspection tokens encourage excess search?. The platform's motive is to capture surplus, and sometimes the losers are the sellers rather than the shoppers.

The core mismatch is that platforms and consumers keep score differently. A platform ranks search designs by conversion revenue. Consumers also count their own time and effort. One model shows this gap means a platform will want agentic search before consumers rationally should switch to it, which explains why people hold on to manual search longer than platforms expect Why do consumers delay adopting agentic search?. The business model adds pressure too. A former Visa Chief Data Officer warns that ad-funded shopping agents have built-in reasons to favor paid placements, however well they are engineered Do advertising-funded shopping agents favor paid product placements?. Ad revenue isn't safe either. If agents fulfill needs in zero-click transactions, search ad income shrinks and the money moves toward social platforms that create wants Do AI shopping agents destroy search ad revenue while leaving influencer marketing intact?. That squeeze gives platforms a reason to find new ways to monetize the agent itself. Services will compete for an agent's attention the way they once competed for clicks Will agents compete for attention just like users do?.

The less obvious finding is that harm compounds without anyone colluding. In simulations, a platform tunes its ranking and sellers tune their product descriptions, each on its own. Because both exploit the same AI cognitive biases, consumer harm more than doubles compared with either one acting alone Do platforms and sellers exploit AI shoppers together more effectively? Can platform ranking and seller descriptions separately harm AI shoppers?. Those biases are real. In marketplace tests, LLM agents favor the first offer they see over better ones by 10–30x and get close to optimal outcomes only when search conditions are ideal Do LLM agents achieve fair marketplace outcomes when given ideal search?. Cheaper search doesn't fix this. One market-design paper argues that near-free agent search, combined with lookalike descriptions and low-quality sellers entering the market, can leave a market worse off than before Can cheap agent search actually harm shopping markets?.

The conflict isn't unavoidable. A well-designed recommendation algorithm can match the welfare of an ideal market mechanism by treating sellers' posted prices as signals of their costs Can an algorithm recommend products without matching an ideal mechanism?. Its advantage over ordinary price competition fades as more sellers enter, though, so the platform's power to shape outcomes matters most in thin markets Does algorithmic recommendation lose its edge as sellers compete more?. The most practical finding is how cheap one fix can be. Agents employed by a platform go easier on sponsored products. Adding one sentence to the agent's prompt that names the consumer as the beneficiary brought its scrutiny of sponsorship back to the level of an agent working for the consumer Can naming the consumer fix AI agent sponsorship bias?. Part of the conflict, then, comes down to whose interests the agent is told to serve.