Spotify says it has removed more than 50M tracks it classifies as AI-generated slop or spam from user recommendations over the past year.
The company announced the figure Tuesday alongside changes to how its recommendation system weighs human-made music. Editorial playlists now apply the same filter, with programmers told not to schedule content identified as slop or spam.
Spotify reported a marked drop in listener complaints about this material surfacing in recommendations after the changes went live.
Detection is the hard part
Nothing in the announcement explains how the company separates generated tracks from human ones. That gap matters. Detection tools have repeatedly produced false positives against legitimate producers who use AI-assisted tools, and independent artists have little recourse when an upload disappears.
The streaming economics push the other way too. Generated tracks cost almost nothing to produce and can be uploaded in volume, which makes them attractive to anyone gaming per-stream payouts. Deezer took a related approach earlier this month by letting listeners hide AI-made music from their feeds outright.
Roughly 50M is a large number in a catalogue that runs past 100M recordings, though Spotify did not say how the removed tracks were counted or whether any remain available on demand rather than merely demoted.
The wider read is that recommendation is becoming the enforcement layer. Platforms cannot police uploads at the volume generated music arrives in, so they filter at the point of delivery instead, and the listener only ever sees what survives the sort.
