Deezer Says 44% of New Uploads Are AI-Generated — and Most Streams Are Bots

Music streaming is how most people listen now. It’s convenient, cheap, and you don’t have to deal with physical media. But it also means AI-generated tracks can slip into your playlists without you noticing. Most platforms don’t bother labeling this stuff, but Deezer has been working on detection tech for a while. Their latest update is sobering: nearly half of all new uploads are AI-made, and most of the “listeners” are AI too.

AI music doesn’t get as much attention as image or text generation, partly because it’s harder to spot. With the right prompts, an AI track can sound like generic, over-produced human music. Deezer ran a survey where listeners heard three songs — two AI, one human. 97 percent couldn’t tell the difference. That’s not a typo. Almost nobody could pick out the human track.

Deezer says it’s built detection tech that identifies AI uploads, and they’re one of the few services that actually labels them. The numbers are wild: 44 percent of new uploads are AI, which works out to 75,000 tracks every single day. They license this detection to third parties and claim a false positive rate under 0.01 percent. I’d take that with a grain of salt until independent audits come in, but it’s still impressive if true.

The bigger issue is that most streams on these AI tracks aren’t real people. Bots are inflating numbers, which messes with royalty payouts and chart rankings. Deezer hasn’t shared the exact bot percentage, but they say it’s the majority. So we’re in a situation where AI generates music, AI listens to it, and real artists get squeezed out of the revenue pool.

I’ve been watching this space for years, and it’s getting weird fast. The music industry already had a bot problem before generative AI blew up. Now it’s compounding. Deezer’s approach is better than nothing, but labeling alone won’t fix the structural issues. Streaming platforms need to rethink how they verify both uploads and plays. Otherwise, we’re heading toward a system where humans are just bystanders in their own music ecosystem.