Spotify removes over 50 million AI-generated tracks from recommendations

To detect such tracks, the service considers several indicators simultaneously: listener reactions and feedback, metadata, upload patterns, and results from third-party AI content detectors. The company explains that detectors can make mistakes, which is why there are many signals: this reduces the risk of mistakenly removing real artists from recommendations. In addition, Spotify has completely removed over 75 million tracks recognized as spam from the platform — according to 3DNews, this took place last year.

Artists who are not real people and perform under an artificially created persona receive an "AI Persona" tag. By default, their music does not appear in algorithmic recommendations or editorial playlists, but subscribers of such an artist will continue to see their tracks in selections. The service has not introduced mandatory labeling for all music made with AI; however, musicians can specify how they used neural networks: for vocals, instruments, lyrics, or production. This information is displayed in the Song Credits section of the mobile app.

The company also reported the verification of more than 300,000 real musicians and the rollout of an Artist Profile Protection tool, which will allow creators to monitor releases under their name and remove erroneous associations (according to Habr, it will be released before the end of 2026). Since 2024, Spotify has had a monetization threshold of 1,000 streams; tracks likely created by AI, according to the service, fail to reach it ten times more often than others.