Beyond the top hits: How generative AI recommendations reshape consumer choice in music streaming services

Abstract

As music catalogs deepen and choice overload grows, recommendation systems increasingly shape what listeners notice and replay. Within this attention economy, generative AI recommendations translate plain-language prompts into intent-matched playlists, yet evidence on their market-level impact remains limited. This study leverages track-level panel data to examine how such a feature reshapes consumption on a large music-streaming platform by comparing outcomes before and after its introduction. We find that streams shift within the observed distribution of popular tracks: streams for higher-ranked tracks decline, whereas streams for lower-ranked tracks increase, consistent with attention fragmentation at the top and greater dispersion toward less prominent tracks. We also examine entry and retention and show that, after introduction, new tracks are more likely to enter active listening and to persist longer once surfaced. Overall, the results suggest that lowering exploration frictions through generative AI recommendations reallocates attention from a narrow set of blockbusters toward a broader share of the catalog. The findings clarify when recommendations diversify consumption by design and offer theoretical and managerial implications.