Zero. That is the approximate number of major streaming platforms that actually bought Deezer’s AI detection technology when it was offered as a professional B2B service. Instead of a windfall from industry partners, Deezer is now pivoting the tech into a consumer-facing tool that allows users to scan their playlists on other services to find the fakes.
Deezer is betting heavily on the “purity” angle. It is a bold move to position themselves as the digital sommelier of the streaming world, pointing out that the vintage you are enjoying is actually synthetic grape juice produced in a lab. For the average listener, a catchy song is a catchy song regardless of whether it was written by a tortured artist in a basement or a GPU cluster in Oregon. But for the audiophile and the developer, the distinction matters because it is fundamentally about provenance and the value of human effort. Does this actually work? If the detection relies on spectral fingerprints or specific artifacts left by current diffusion models, it is a losing battle. Like a food inspector at a carnival, Deezer is trying to spot the fake ingredients in a world where the recipes change every single day.
The technical reality is that AI detection is a treadmill. Suno and Udio are moving faster than any detection API ever could. The moment a detector becomes effective, the generators optimize to bypass it. It is the same cat-and-mouse game we have seen with AI-generated text—once the patterns are known, they are erased. This means any tool that claims to “detect” AI is essentially a snapshot of what AI looked like six months ago. (Or maybe not—see below.) If Deezer has found a way to identify a fundamental mathematical property of synthetic audio that cannot be masked, they have something real. But in this field, “something real” usually has a shelf life of about one update cycle.
Then there is the friction of the actual user experience. But why would a user actually do this? No one is going to migrate their entire library to Deezer just to run a purity test on their favorite 2010s pop hits. The manual labor required to move thousands of tracks between services is a known headache that most people simply avoid. However, the most telling part of the report from The Verge is that Deezer offered this tech to other platforms and got ghosted. From a cold business perspective, one could argue that the bigger players have little incentive to purge this content. Synthetic tracks don’t demand royalties from humans; if a platform can shift a percentage of its plays toward AI-generated content that it owns or licenses cheaply, the margins improve.
Still, the sheer volume of AI slop hitting the platforms is reaching a breaking point where it degrades the product for everyone. The noise floor is rising, and the “discovery” algorithms are starting to choke on low-effort, high-volume synthetic uploads. This creates a vacuum that Deezer is trying to fill by acting as the arbiter of authenticity. By Q3, we will see the first major “Verified Human” certification badge appear on a top-tier streaming platform to combat the flood of synthetic content. Until then, we are left with a tool that tells us our favorite new lo-fi beat was made by a machine—information that is intellectually interesting but emotionally useless.
A noble effort to save the soul of music that will likely be ignored by the people who actually control the playlists.
