Work

Beyond Retrieval: Generative Algorithms in Recommendation and Search

Type: article
Abstract: Generative artificial intelligence is transforming how digital platforms connect users with information, products, and services, a task long performed by retrieval-based recommentaion and search algorithms. This shift raises a central question: how does deploying generative algorithms affect user engagement? Combining China’s mandatory algorithm filing records with app-level monthly active user data, we trace the diffusion of generative algorithms and estimate their engagement effects using staggered and stacked difference-in-differences designs. Generative recommendation and search algorithms diffuse rapidly across app categories and produce larger increases in monthly active users than their conventional counterparts. Their adoption also reduces competitors’ user engagement, indicating that generative capabilities reshape the allocation of attention within platform markets. These effects arise across both demand-formation and demand-fulfillment contexts. Our findings show that generative algorithms move digital intermediation beyond information retrieval, from ranking existing items to constructing answers.
Language: English
Open Access status: green
OpenAlex ID: W7222990266
Cites: 13
Cited by:
Related to: 10