Truace
Subscribe

Spotify removes 50M AI-generated tracks from recommendations

Spotify has removed over 50 million AI-generated slop and spam tracks from its recommendations, aiming to prioritize music created by human artists. Spotify has improved its recommendation system based on listener feedback and engagement. The company found that tracks users skipped, ignored, or reported were often identified as slop or spam. It now uses a combination of factors such as upload pat

By the Truvace desk1 min read1 sourceJournalismTRV-2026-2137· cite & verify
Spotify removes 50M AI-generated tracks from recommendations

In brief

Spotify has taken tens of millions of AI-generated tracks that it classes as slop or spam out of its recommendation feeds. The stated purpose is to steer listeners toward music made by human artists.

The move shows a streaming platform responding to low-quality synthetic music by adjusting what it promotes. The supplied text is truncated, so the full detection method, any effect on artists' earnings, and how well the filtering works remain unclear.

Main points

  1. The reported scale is more than 50 million AI-generated tracks removed from recommendations.
  2. Spotify improved its recommendation system using listener feedback and engagement signals.
  3. Tracks that users skipped, ignored, or reported were often identified as slop or spam.
  4. The company uses a combination of factors, including upload patterns, in its detection approach.

The gain

Spotify says it removed over 50 million AI-generated slop and spam tracks from recommendations, aiming to give human-made music more prominence.

The rundown

The change targets recommendations rather than described as a wholesale ban: the tracks were removed from the recommendation system, with the goal of prioritizing human-created music.

Spotify relied on listener behavior as a signal. Skipped, ignored, or reported tracks were frequently found to be slop or spam, which fed into the improved system.

The article then begins to describe a combination of factors such as upload patterns, but the text cuts off, so the remaining detection criteria are not available.

Sources

  1. JournalismNewsBytes2026-10-21

The debate