Spotify deletes over 50 million AI‑created songs to prioritize human‑made tracks in recommendations

Spotify said it is undertaking a massive purge of its recommendation system, removing more than 50 million tracks it identified as AI‑generated or spam‑like. The action, outlined in a blog post and covered by Gizmodo, aims to make sure listeners mainly receive suggestions from human creators rather than automated uploads.

The streaming service explained that the sweep focused on songs flagged by its internal detection tools for missing genuine human input, containing repetitive metadata, or showing other signs of low‑quality, algorithm‑produced content. Although the precise criteria are kept confidential, Spotify indicated the cleanup is part of a wider effort to boost the relevance of its “Discover Weekly” and “Release Radar” playlists.

Spotify’s recommendation algorithm has traditionally combined collaborative filtering, natural‑language processing of track metadata, and analysis of user listening habits. Recently, the platform has observed a rise in uploads that seem to be mass‑generated by AI models, often bearing generic titles and repetitive structures. Critics argue that such material can water down the listening experience and make it tougher for up‑and‑coming artists to be discovered.

Analysts in the industry point out that the move mirrors a growing unease throughout the music‑streaming world about generative AI’s effect on creative ecosystems. While AI‑assisted composition tools can support musicians, an unchecked flood of fully synthetic tracks raises issues around copyright, royalties, and the authenticity of curated playlists. By weeding out what it calls “spam,” Spotify seeks to safeguard its catalog’s integrity and protect licensed artists’ revenue.

Most artists and rights‑holders have welcomed the measure, seeing it as a guard against algorithmic exploitation. A spokesperson for the Recording Industry Association of America (RIAA) said the cleanup fits within broader attempts to fight “digital noise” that can erode the value of recorded music. Nonetheless, some independent creators who experiment with AI‑driven music voiced worries that the new filters could unintentionally block legitimate experimental works.

Spotify has not revealed what share of its total library the 50 million tracks constitute, but internal estimates suggest it is a modest portion of the service’s more than 80 million songs. The company intends to keep fine‑tuning its detection tools and will track the effect on user‑engagement metrics in the weeks ahead.

Looking forward, the streaming giant said it will invest in “transparent, human‑centric recommendation models” and will give artists clearer guidance on acceptable metadata practices. The rollout could act as a bellwether for other platforms wrestling with the balance between AI innovation and the preservation of human artistry in the digital era.