Nearly Half of Music Uploaded to Deezer Is AI-Generated, But Almost Nobody Is Listening

Something quietly extraordinary is happening in the pipes of the music industry. Deezer, the French streaming platform with roughly 10 million active subscribers, has revealed that 44% of all songs uploaded to its platform every single day are now AI-generated. That figure alone would have seemed like science fiction five years ago. What makes it stranger still is the accompanying detail: almost no one is actually listening to any of it.

Consumption of AI-generated music on Deezer sits between just 1 and 3% of total streams. The math here is jarring. Nearly half the content arriving at the platform's door is synthetic, yet it accounts for a vanishingly small share of actual listening behavior. This is not a story about AI music taking over the charts. It is a story about a different kind of pressure entirely, one that has less to do with taste and everything to do with economics.

Deezer's own data offers the most revealing detail in this whole picture: 85% of AI-generated streams on the platform are detected as fraudulent and subsequently demonetized. That means the overwhelming majority of AI music being uploaded is not an artistic experiment or even a commercial gamble. It is an attempt to game royalty systems at industrial scale.

Streaming royalties, however modest per play, are essentially a numbers game. If you can generate thousands of tracks cheaply and then use bots or coordinated streaming farms to rack up plays, the per-stream fractions add up. The music itself is almost incidental. It is a vessel for extracting micro-payments from a pool that every legitimate artist also draws from. This is sometimes called "stream fraud" or "royalty fraud," and it has been a known problem in the industry for years, but the arrival of accessible AI music generation tools has turbocharged the supply side of the equation in ways that manual fraud operations simply could not match.

The broader streaming ecosystem has been grappling with this for some time. In 2023, Universal Music Group and other major labels pushed Spotify and other platforms to take stronger action against AI-generated content flooding royalty pools. Spotify subsequently removed tens of thousands of tracks linked to a company called Boomy, whose AI-generated songs had been streamed fraudulently. Deezer's disclosure suggests the problem has not gone away. If anything, the infrastructure for generating and uploading synthetic content has become cheaper and faster.

The second-order consequence worth watching here is subtler than fraud. When nearly half of daily uploads are synthetic, the sheer volume of content on a platform begins to change the nature of the platform itself. Music catalogs on major streaming services already run into the hundreds of millions of tracks. Spotify reported over 100 million songs on its platform in 2023. If AI generation continues at this pace across the industry, that number could balloon in ways that make meaningful discovery increasingly difficult.

This creates a feedback loop with real consequences for human artists. Royalty pools are not infinite. They are divided among all monetized streams. As fraudulent AI content gets demonetized, it does not directly steal from artists in that moment, but the infrastructure costs of detecting, flagging, and removing it are real. More importantly, the legitimate AI-generated content that does get through, the 15% of AI streams Deezer is not flagging as fraudulent, does compete directly in that pool. A smaller share of a pool divided among more participants means less for everyone who plays by the rules.

There is also a trust dimension. Platforms that become known as repositories of synthetic filler risk eroding the listener relationship that makes them valuable in the first place. Deezer's willingness to publish these numbers is notable precisely because it signals the company understands that transparency may be a competitive asset as listeners grow more aware of what is flooding the system.

The deeper question is whether detection can keep pace with generation. Right now, Deezer claims an 85% catch rate on fraudulent AI streams, which sounds impressive until you consider that the volume of AI uploads is itself growing. A constant percentage applied to an accelerating base is not a stable equilibrium. It is a race, and the tools on both sides of it are getting faster at roughly the same time.