Spotify removes 50 million AI slop and spam tracks from recommendations

Spotify says it has removed more than 50 million songs it classifies as AI-generated slop or spam from user recommendations over the past year. According to The Verge, the company disclosed the figure in an announcement that also described changes to its recommendation system designed to highlight more music made by human artists.

The move is one of the clearest platform-level responses so far to the flood of AI-generated audio hitting streaming services. Spotify's editorial teams are being held to the same rule. "Our editors hold the same standard and don't program content identified as slop or spam," the company said, adding that since the changes were made it has "seen a marked decrease in listener concerns about this content showing up in recommendations."

How Spotify spots AI slop

According to The Verge, Spotify relies on a combination of listener engagement and feedback, upload patterns, metadata and third-party AI detection tools to judge whether a track may be AI-generated slop. No single signal is definitive, which points to how difficult it remains to reliably separate machine-made music from human-made music at scale.

The recommendation purge also continues a broader cleanup. Spotify said last year that it took down more than 75 million songs identified as spam, The Verge reports.

A badge instead of a blanket label

Spotify has stopped short of requiring labels on all AI tracks. Instead it launched an "AI Persona" badge for artists that use an AI identity and don't represent a real person, and music from these personas will not be recommended at all. The effect is to push such content out of discovery while still allowing it on the platform.

That leaves Spotify in the middle of the pack among music services. As The Verge notes, YouTube Music puts labels on AI songs, Deezer gives users an option to filter AI slop out, and Bandcamp has banned AI music entirely.

The company is also not uniformly hostile to AI in music: it is partnering with Universal Music Group and the independent music publisher Believe to launch a tool for creating AI music remixes. Its target appears to be mass-produced spam rather than the technology itself.

Why it matters

The headline number concerns recommendations, not the catalogue as a whole. AI-generated tracks can still live on Spotify; what they lose is algorithmic reach, which is precisely what made mass-uploading them worthwhile. For listeners, that should mean cleaner playlists and autoplay queues.

It also sketches where AI content moderation is heading. Detection is heuristic and imperfect, platform policies vary widely, and the sheer volume involved, tens of millions of tracks, shows how cheap generation has become. If recommendation systems become the main choke point for AI slop, the companies operating those systems, rather than copyright law or disclosure rules, will largely decide what audiences get to hear.