Signal Audit of Commercial Generative Music Systems Finds Measurable Homogenization Across Genres Including K-pop

Summary

A study by Zoe Slendebroek and Danae Metaxa audited whether large-scale generative music systems show measurable musical homogenization compared with human-produced music. The study also offers a justice-centered account of why that homogenization matters. The researchers audited two commercially deployed systems, Suno and Lyria 3, across four genres, namely Afrobeats, K-pop, Dance Pop and Heavy Metal. For each system and genre they generated 100 tracks and compared them against corpora of human-produced music in the same genres. Because K-pop is included, the study has a direct touchpoint with Korean popular culture. It examines whether AI-generated music flattens genre-specific musical diversity across a wide range of styles worldwide.