Qobuz Tags AI Music and Drops a Bombshell: 60% of AI Track Streams Are Fraudulent

If you run a label, manage a roster, or move catalog through a distributor, here is the number that should stop you cold: nearly two out of every three streams on AI-flagged tracks are fake. That is not a projection. That is what one streaming service is now reporting from its own systems.

For anyone who has spent the last two years being told that AI music and streaming fraud are separate conversations, this is the moment the two problems officially fused into one.

Qobuz, the France-based hi-res service, has started displaying a tag in its app that alerts users when a release has been identified as AI-generated. But the transparency headline buries the real story. As Music Business Worldwide reported, Qobuz says 60% of streams from tracks it has flagged as AI-generated are currently deemed fraudulent by its anti-fraud systems. Those streams get excluded from royalty payouts.

Read that again. This is not a company estimating how much AI music might be sketchy. This is a DSP saying that when it looks at the AI pile, the majority of the listening is manufactured.

What Qobuz actually did

The tag appears at release level, flagging albums that contain content Qobuz identified as AI-generated. The detection mechanism was built in-house, per Qobuz, and it builds on the AI Charter the company first published in February 2026, now updated to a second version dated September 2026.

Qobuz also went further than labeling. It removed more than a third of the AI-identified albums that had zero listening activity from its search engine. Those de-indexed albums cover four million tracks, the company said. They stay findable if you search the exact title, but they no longer surface in discovery.

AI-flagged content is also kept out of every human-curated corner of the app (Our Selections, New Releases, Qobuzissimes, Albums of the Week, the playlists) and excluded from algorithmic recommendations. Qobuz credits that human-led approach for why AI tracks are only 0.38% of its total streams, a figure that should make every volume-driven platform squirm by comparison.

Why the fraud number matters more than the label

Labeling AI music is now table stakes. Deezer, TIDAL, and Apple Music have all committed to some version of it. What Qobuz added is the quantified link that the industry has been circling for two years without proof.

The 60% figure lands in a broader pattern. Music Business Worldwide noted that Deezer said in July it was receiving roughly 90,000 fully AI-generated tracks every day in June, and that on peak days that content passed 50% of all new music delivered. Deezer also reported that up to 85% of streams on such tracks were fraudulent in 2025.

So now you have two services, using different methods, arriving at the same conclusion from opposite ends: the money moving through AI catalog is overwhelmingly fake. When the fraud rate on a content category sits between 60% and 85%, that category is not a music problem. It is a laundering problem wearing a melody.

That is the pressure point for Spotify and Apple Music. At their scale, even a small percentage of AI content represents an enormous absolute volume, and if the fraud rates track anywhere near what Qobuz and Deezer are seeing, the leakage from the royalty pool is not theoretical. It is being paid out right now, to somebody, in place of your artists.

The volume fight underneath all of it

The uncomfortable truth is that fraud thrives on volume, and volume is the business model for a lot of the distribution chain. Universal's recent suit against DistroKid made the point bluntly, citing nearly 12 million tracks uploaded through that company in six months. (We broke down the stakes in that case here.)

Independent think-tank ORCA wants to attack the flood at the source. Executive director Patrick Clifton argued in a blog post, as Music Ally covered, that it is time for upload fees across the digital supply chain to disincentivize bad actors, because current approaches have not worked. His proposal: a per-track fee on new recordings, calibrated to avoid punishing legitimate artists, with DSPs collecting the fees and folding them back into royalty pools.

You can debate the mechanism. What you cannot debate anymore is the premise. Free, infinite uploading plus AI generation plus bot streaming is not a bug in the system. It is the system, and Qobuz just put a number on the exit wound.

Qobuz's own charter admits the limit here: no reliable solution exists today to identify all AI content with certainty, especially when a human touched part of it. Fine. But the industry has spent long enough hiding behind that caveat. The tools do not need to be perfect to prove the pattern.

The smallest hi-res service in the room just did the math the giants did not want on paper. The question is no longer whether AI content and fraud are the same fight. It is which platform admits it next, and how much of your money walked out the door while everyone waited.

Sources

• Music Business Worldwide
• Music Ally