Spotify said Tuesday that it had filtered about 50 million songs identified as spam or low-quality AI-generated content out of recommendations for its users. The measure changes what music the service suggests: being left out of those recommendations does not mean a song has been removed from the catalog.
The company explained that it adjusted its system to give more prominence to music by human artists. Its editors also exclude content identified as spam or low quality from programming. The changes therefore affect both editorial selection and automated suggestions.
How it decides what content to filter
Detection combines listener engagement and feedback with song-upload patterns, metadata, and external tools for identifying AI. Among the signals Spotify reviewed were tracks that users skipped, ignored, or marked as music they did not want to hear.
Spotify says complaints about this content appearing in recommendations declined after the changes. That assessment comes from the company itself; it does not prove that the filter correctly detects all AI-generated music.
Another measure is the “AI Persona” badge, intended for profiles with an artificial identity that does not represent a real person. By default, these identities are excluded from editorial and algorithmic recommendations, though exceptions exist—for example, when a user follows that profile.
The platform has also added a feature for viewing AI contributions in song credits within the app. This does not mean all tracks generated with artificial intelligence are required to carry a label: the artificial-identity badge and the credits information are separate measures.
Image: The facade bearing the Spotify logo in a file photo from September 2007. Erik Stattin / Wikimedia Commons, CC BY-SA 2.0.
