Prices on online marketplaces such as Amazon are often set by algorithms that adjust them based on what rivals are charging and how much sellers bid for visibility in search results — a cost that feeds into pricing.
Their growing use has sparked concern over digital price fixing, where algorithms could move together to push prices higher at consumers’ expense. But that is not always how these systems behave, according to new research from Wharton marketing professor Ron Berman and Hangcheng Zhao of Rutgers Business School.
Drawing on data from nearly 2,400 product searches conducted repeatedly on Amazon, the study finds that prices fall in categories where algorithmic pricing is widespread and shoppers consider fewer options before buying.
“A lot of research assumes this is bad for consumers, that algorithms collude on prices and weaken competition,” Berman said. “Our paper shows that is not always the case. Consumers benefit in many cases.”
How Do AI Pricing Algorithms Really Work?
The research, published as a working paper on SSRN, examines online marketplaces where sellers not only set prices but also vie for visibility through paid placements — a model used by platforms from Amazon to eBay and Expedia. Merchants must decide both what to charge and what to spend to secure a spot at the top of search results.
Around a third of sellers on Amazon, the authors estimate, rely on such pricing systems. But instead of driving prices up, the algorithms end up moving together to push advertising bids lower. That cuts merchants’ costs, feeding through into lower prices for buyers and drawing more of them to the platform.
That, in turn, reflects a quirk of how online ads are sold. In auction systems, sellers often overpay for prominent placement — a dynamic known as the “winner’s curse.” The algorithms, the study suggests, learn to avoid that.
The effects do not stop there. Lower prices fuel sales, lifting platform commission revenues enough to more than offset the drop in advertising income, the paper shows. Amazon takes a percentage cut on each sale.
“A lot of research assumes this is bad for consumers, that algorithms collude on prices and weaken competition. Our paper shows that is not always the case.”–Ron Berman
AI Pricing Still Depends on Shopper Behavior
The study was based on more than 2 million products and over 1 million daily prices tracked on Amazon, but the authors say the results extend to similar online marketplaces.
However, the effect is not uniform. It depends on how shoppers behave. When buyers compare fewer products, algorithms push prices below where they would otherwise be. When they compare more, pricing climbs above that level, the study shows.
The difference comes down to how people actually shop. On Amazon, some consumers comb through dozens of options while others click the first results they see and buy. Shoppers typically scan just 15 products, and fewer than a quarter ever make it to items in the middle of the page, according to the study.
“It depends on the product,” Berman said. “For things like toothpaste or toilet paper, advertising matters more. For others, like electronics, people search and compare much more.”
That distinction is central to the policy debate.
“If [marketplaces] offer these algorithms, regulators may say you are helping sellers collude without them talking to each other.”–Ron Berman
Why Regulators Are Concerned About Collusion
Algorithmic pricing has become a flashpoint for regulators who worry these systems could learn to move together to push prices higher and harm consumers. The risk has drawn scrutiny from regulators, including the U.S. Federal Trade Commission, which has said that “price-fixing by algorithm is still price-fixing.”
Recent cases show why regulators are concerned. The FTC sued Amazon in 2023, alleging the company used a secret algorithm to pinpoint products rivals tended to match, then raise prices and lock in those increases once competitors followed. The case remains ongoing, and Amazon has rejected the allegations.
Berman points out that AI raises a unique concern for regulators: Even without speaking to each other, sellers using pricing algorithms can end up moving in lockstep. Collusion usually implies firms explicitly agreeing to fix prices.
“Marketplaces are under heavy scrutiny,” he said. “If they offer these algorithms, regulators may say you are helping sellers collude without them talking to each other.”
