Spotify has kept more than 50 million tracks it classifies as AI slop or spam out of its recommendations so far in 2026, the company said in a newsroom post on Tuesday, October 20, 2026, marking a year since it set out rules against spam, impersonation and deception in the AI era. Those tracks remain on the service, Variety reports, but listeners have to search for them because they are no longer surfaced by algorithms or editors.
The figure comes on top of more than 75 million spam tracks that Spotify says it has deleted entirely for serious or repeated breaches of its policies. NBC News, which also covered the announcement, describes those removals as having taken place last year. Together, the numbers give a rare sense of how much machine-made audio is being pushed onto streaming platforms.
Spotify says the change started with listener behaviour. When it studied tracks that people skipped, ignored or explicitly rejected, the songs overwhelmingly matched content it had already flagged. Over the last half-year it has repeatedly adjusted its recommendation system and says complaints about such content in recommendations have dropped noticeably. Its playlist editors follow the same rule.
Detection relies on several signals rather than one: engagement and feedback, upload patterns, metadata and results from several outside AI-detection tools. Spotify openly acknowledges that AI detection has limits and will get harder as generative models improve, and it says it is also watching for tactics such as mass uploads, duplicates, search manipulation, spam playlist networks and very short tracks.
The post also updates tools introduced over the past year. AI Credits, launched in April, let artists disclose which parts of a song used AI, and Spotify says it now receives tens of thousands of such credits each day. More than 300,000 artists carry a Verified by Spotify badge, over 70% of them independent. Thousands of AI Persona profiles are labeled and excluded from recommendations unless a listener follows them. A beta called Artist Profile Protection, which lets artists approve releases credited to them, is due to reach every artist on the platform before 2026 ends. Spotify adds that likely AI-generated tracks are ten times more likely to fall under its 1,000-stream royalty threshold.
The stakes are financial as well as cultural. Streaming royalties are paid from a shared pool, so a flood of synthetic tracks can shrink every legitimate artist's share. NBC quotes Morgan Hayduk, chief executive of the fraud-detection firm Beatdapp, making that point. Variety notes that YouTube Music, Apple Music and Tidal have also introduced labels for AI-made music.
Spotify did not say how it defines slop in borderline cases, how often legitimate songs are wrongly flagged, or how many AI-assisted tracks it still recommends. The numbers are the company's own and have not been audited.