Tweaking search rankings boosted purchases of platform's own products by 34 percentage points, Fair Trade Commission finds

Algorithm-driven self-preferencing by online platforms can significantly distort consumers' product choices, according to new research.

Simply placing a platform's own products at the top of search results sharply increased purchase rates, the study found, while disclosure measures such as self-preferencing labels and ranking-criteria notices proved insufficient to correct the distortion.

The Korea Fair Trade Commission announced Saturday it is publishing a report titled "An Experimental Study on Consumer Behavior Regarding Algorithm-Based Self-Preferencing by Platforms," based on a randomized controlled trial conducted with 3,072 consumers.

The experiment used a virtual shopping mall replicating a real online shopping interface to analyze consumers' search and purchase behavior. Consumers were found to rely heavily on the search rankings the platform presented.

Purchases were concentrated at the top of search results: 51.7 percent of all purchases came from the top five products, and 94.6 percent were completed within the first page. Only 25.2 percent of consumers changed the default sort order, and 83.8 percent did not use the filter function at all, confirming a strong tendency to follow the platform's default rankings.

Against this backdrop, the research tested the self-preferencing effect in a second showing by placing a platform's own copycat product — identical to a lower-ranked item but priced 10 percent higher — at the top of search results. The purchase rate for that product rose from 1 percent before self-preferencing to 35 percent after, an increase of about 34 percentage points.

The purchase rate for the previously top-ranked competing product fell from 52 percent to 20 percent, a drop of about 32 percentage points. This suggests consumers treat search rankings as a signal of product quality or relevance, meaning simple ranking manipulation alone can substantially alter purchasing decisions.

Disclosure-based remedies designed to reduce distorted choices showed limited effectiveness. Labels attached to self-preferenced products actually increased the purchase rate for those products by about 4.5 percentage points and reduced consumers' active search behavior. Only 10.7 percent of consumers checked the ranking-criteria disclosure.

However, among the smaller group of consumers who did check the disclosure, the purchase rate for self-preferenced products fell by about 18.4 percentage points.

Consumers who bought the higher-priced self-preferenced products tended not to perceive the outcome as a loss. Their satisfaction with the purchase and their trust in the rankings actually increased, confirming a structural pattern that makes it difficult for consumers to recognize on their own that an algorithm has distorted their choices.

The Fair Trade Commission described the study as its first experimental research to establish the causal effect of algorithm-based platform self-preferencing on consumer choice.

"Because the confidentiality and opacity of algorithms make it difficult to prove causation between conduct and market outcomes in platform markets, experimental methodologies such as randomized controlled trials are expected to serve as a useful analytical tool to complement competition policy research and law enforcement," the commission said.

It added that it would "actively employ a range of economic analysis methodologies — including experimental research, econometric analysis, and behavioral economics — to address new competition issues in digital markets."