Spotify Removed 75 Million AI-Generated Tracks in a Year

Spotify has confirmed that it removed more than 75 million tracks from its catalog over the past year after flagging them as AI-generated spam. The move is a response to a wave of mass-uploaded music with a single purpose: gaming the recommendation system for as many streams as possible and cashing in on royalties.

The scale of the problem

The 75 million figure is striking, but it needs to be weighed against how fast Spotify's catalog is growing. About 100,000 new recordings are uploaded to the platform every day, while rival Deezer sees as many as 170,000 new tracks daily, an increasing share of which is music created entirely by generative models. Deezer says the number of fully AI-generated tracks it receives has grown from 10,000 a day in January to 30,000 now.

The issue isn't the technology itself but how it's being used. Spotify doesn't ban music made with the help of AI, as long as the creator holds rights to the material used, doesn't clone someone else's voice without consent, and shows genuine creative input. What the platform is cracking down on is mass, automated uploads of low-quality recordings, duplicates with swapped metadata, and tracks deliberately shortened to just clear the threshold at which a stream starts earning royalties.

New rules

The company announced three pillars of its policy on AI music. The first is an outright ban on unauthorized cloning of an artist's voice, treated as a violation of the creator's identity. The second is a new anti-spam filter that detects mass uploads, duplicates, SEO manipulation in track descriptions, and artificially shortened songs, excluding them from recommendations. The third pillar is a partnership with the organization DDEX on a metadata disclosure standard that would show how much AI was involved in making a recording, from vocals through instrumentation to post-production.

Behind the effort is a dedicated team Spotify set up specifically to detect and remove this kind of content. Sam Duboff, the company's head of artist and industry marketing and platform policy, described the removed content bluntly.

Low-effort, mass-produced, mass-uploaded, that's junk created by artificial intelligence - Sam Duboff, head of artist and industry marketing and platform policy, Spotify

Why it pays off for scammers

The scale of abuse is rising along with the money flowing through streaming platforms. Spotify notes that its payouts to the music industry grew from $1 billion in 2014 to $10 billion in 2024. The bigger the royalty pool, the greater the temptation to flood the catalog with cheap, mass-generated music and skim streams from playlists and algorithmic recommendations.

Bot-driven traffic makes the problem worse. Deezer, the first platform to deploy a synthetic-music detector, estimates that as much as 70 percent of streams on fully AI-generated tracks come from fake activity rather than real listeners. The issue drew wide attention after the case of The Velvet Sundown, a band that racked up more than a million streams on Spotify before it came out that its music and the members' images had all been generated by AI.

Industry reaction

Universal Music Group praised Spotify's new rules, calling them consistent with the label's own artist-first principles. Rival platforms are moving in a similar direction: Tidal has introduced a policy of labeling AI tracks and removing profiles impersonating real artists, while Deezer is developing its own synthetic-music detector and has been publicly warning about the scale of the phenomenon.

For Polish artists and publishers, this means the streaming market is tightening control over what makes it into catalogs and playlists. Artists legitimately using AI tools should expect to face a requirement in the coming months to disclose that use in track metadata, under the standard being developed with DDEX. For listeners, the practical change will be fewer random, low-quality recordings impersonating real performers showing up in recommendations and playlists.

Spotify says this isn't the end of the effort, the spam filter will be rolled out gradually over the coming months, and the team dedicated to detecting abuse is meant to remain a permanent part of the platform's moderation.