Spotify Removes 50 Million AI Tracks From Recommendations
Spotify states it has removed more than 50 million AI slop and spam tracks from its recommendations this year, alongside more than 75 million removed from the platform entirely.
- Out of recommendations: more than 50 million tracks, this year
- Off the platform: more than 75 million, for serious or repeated violations
- AI personas: labelled, and barred from editorial and algorithmic placement
- Verified by Spotify: more than 300,000 creators validated
How Spotify Says It Finds Them
Behaviour rather than audio alone. The company cites instant skip rates, suspicious metadata and high-frequency upload patterns, combined with third-party AI detection, and says it is targeting search manipulation, mass duplicate uploads, artificially truncated tracks and bot-driven playlist networks built to collect royalties.
Truncated tracks and playlist networks are the tell. Both are optimisations against a payout rule rather than attempts to be listened to.
The Rule Doing The Quiet Work
A threshold of 1,000 streams before a track earns anything. Spotify states tracks showing heavy synthetic markers are ten times more likely to fail that test, which makes the economics hostile to volume before any moderation happens.
That is the more durable mechanism. Detection is a contest that improves on both sides; a payout floor simply makes the business case for mass uploading worse.
What It Means For Artists And Rights Holders
Disclosure is being built alongside enforcement. AI Credits, introduced in April, let artists declare synthetic contributions across vocals, lyrics and production, and Spotify states adoption has been rapid.
Verified by Spotify has validated more than 300,000 creators, which the company says account for more than 99 per cent of listener search volume, and an Artist Profile Protection tool rolling out by year end will let artists vet incoming track tags before unauthorised collaborations appear on their pages. The through line is provenance: who made it, and who is allowed to say so.
More ad tech coverage on Ad Tribe.