Deezer has published its latest update on how much AI-generated music is being uploaded to its service. The total now stands at nearly 75,000 tracks a day – 44% of total daily uploads.
As a quick recap, this has increased from around10,000 tracks a day in January 2025, when they accounted for around 10% of daily uploads.
The growth over the past 15 months has been sharp: 20,000 AI tracks a day by April 2025; 30,000 by September; 50,000 by November; and 60,000 by January 2026. Or if you prefer that as a graph:
Let’s talk about human music for a minute. Based on Deezer’s figures, in January 2025 it was receiving around 90,000 non-AI tracks every day – the other 90% of its uploads. This week’s update suggests it’s now around 95,500.
In other words, the amount of human-made tracks being uploaded to Deezer hasn’t grown much over the past 15 months alongside this surge in AI-generated material.
Another trend from Deezer’s data-drops continues though: AI-generated music is still only 1-3% of the service’s total streams, and 85% of those streams are “detected as fraudulent” by Deezer – so are not earning royalties. These two statistics are the same as in its January announcement.
So… what? The key question here is: even if an ever-growing mountain of AI music is being uploaded to streaming services, if it’s a.) not being listened to by humans much at all and b.) if the fraudulent stuff is being detected and demonetised successfully… is it really a problem?
The answers to which might be: a.) AI slop dogging up the DSPs’ servers is a cost problem for them, and an environmental problem for all of us; and b.) we’d be wise not to be too complacent about fraud-detection, even if it’s successful at the moment.
Note, Deezer’s decision to regularly update its scary figures while stressing that its detection tech is demonetising fraud is also a marketing tactic, since it’s trying to license that tech to other companies in the industry. There’s nothing shady in that strategy, but it’s relevant context.
Even so, the transparency is welcome: a running count for the industry on the scale of uploads (big) and listening to those uploads (small). And useful numbers for DSPs to digest when deciding how they should deal with this music – a subject we explored at more length in our recent insight report.
