A purge aimed at synthetic music
Spotify has taken more than 50 million tracks that look like they were made by artificial intelligence out of its curated music recommendations since the start of the year, the company announced Tuesday. The streaming platform framed the move as an effort to prevent machine-made music from pushing human artists aside. In a statement, the company said AI-generated slop and spam have no place in listener recommendations and that it had adjusted its systems to block them and safeguard the listening experience.[S1][S2][S3][S4][S5][S6][S7]
The figure follows a larger cleanup last year, when Spotify removed 75 million tracks for what it described as serious or repeated breaches of its rules. Those rules bar copying an existing creator's work and forbid producing media artificially with technology. The scale of both numbers reflects the difficulty streaming services face in spotting impersonation and deception among tens of millions of tracks uploaded online.[S1][S8]
How listeners helped flag the slop
According to Spotify, subscribers played an unexpected role in the effort. Over the previous six months the company studied patterns in songs that users regularly skipped or passed over, and found that those tracks overlapped consistently with artificially generated audio, sometimes called slop. By pairing that listener feedback with metadata and outside AI-detection tools, Spotify said it was able to retune its internal recommendation system and reduce how often such content appeared.[S1]
The company conceded that AI-detection technology is imperfect and that telling suspicious content apart may get harder as generative AI advances. Spotify said the combined approach has produced a marked decrease in complaints from listeners about AI spam surfacing in their recommendations. It also argued that as AI makes music more plentiful, audiences are likely to place greater value on human artistry.[S1]
Why the royalty math matters
For artists, the stakes go beyond what shows up in a playlist. Spotify pays out partly according to how often a track is played relative to the platform's total streams, so a swollen pool of AI tracks can dilute what each legitimate rights holder collects. Morgan Hayduk, chief executive of Beatdapp, which sells AI- and fraud-detection software to labels and distributors, said that if the denominator grows because of AI slop, the percentage each legitimate rights holder or artist receives is diminished.[S1]
Spotify requires an artist to reach at least 1,000 streams before royalties are paid, a threshold that AI tracks clear far less often than human-made music. The platform has also tried to accommodate artists who use AI as part of their creative process without producing what it would call slop. Earlier this year it introduced an AI credits feature letting musicians disclose which parts of a song's creation involved AI.[S1]
Other transparency measures have arrived alongside it. Spotify attaches an AI Persona badge to the profiles of musicians it believes are machine-generated, and those acts are kept out of recommendation streams unless a listener actively follows them. The company said the mix of audience signals, metadata and detection tools is one of several steps it has taken this year to clarify how AI is used on the service.[S1]
