Music-streaming listeners may soon get a much more direct way to answer a question that has been hanging over recommendation feeds: do you want machine-generated tracks mixed into your music, or not? Deezer says it will add an in-app toggle by the end of the year that can completely filter AI-generated music from a listener’s experience.
It is a notably plain-language solution to a complicated platform problem. Rather than merely marking a track, burying a disclosure in artist information, or asking listeners to decipher whether an upload is synthetic, the proposed setting is intended to let users opt out. Flip the switch, and the service aims to focus the feed on music made by human creators.
That degree of control matters because the issue is no longer hypothetical background noise. Deezer says fully AI-generated uploads have grown from roughly 18% of tracks submitted to its platform last year to more than half. It also says about 90,000 fully generated tracks are uploaded daily, bringing the cumulative total to more than 42 million.
A filter, not simply another label
An AI-generated track, in this context, means music created fully by an artificial-intelligence system rather than a conventional human performance or composition process. The planned Deezer control is designed as a filter: a setting that excludes a category of material from what a user hears. That is substantively different from a label, which identifies or discloses a category but leaves it present in the catalogue and potentially in recommendations.
Deezer chief executive Alexis Lanternier framed the tool as a response to listener demand and as an option rather than a universal ban:
“We are now giving everyone the option to remove this content completely from their streaming experience on Deezer, and focus only on human creators and creations, if they wish.”
The distinction is important. A label asks the listener to make a choice track by track. A filter changes the default pool from which the service can surface music for that listener. In practical terms, someone who wants to avoid AI music would not need to scrutinize every unfamiliar artist or playlist placement before pressing play.
It also means the platform is presenting two listening preferences as legitimate: listeners can continue to encounter AI-generated music if they wish, or can remove it from their experience. That is a more specific intervention than simply informing people that an AI track exists.
Why the volume creates a discovery problem
The core problem is not only whether a single generated song sounds convincing. It is scale. When tens of thousands of tracks can arrive in a day, the catalogue, search results and recommendation systems face a flood of new material. A listener looking for a human musician can end up dealing with an identification problem before they can make a taste judgment.
Deezer’s figures suggest that generated music is now a major share of incoming uploads on its service. More than 50% is a particularly meaningful threshold: it means AI-generated tracks are no longer a small edge case among new submissions. The proposed toggle effectively acknowledges that a portion of the audience does not want algorithmic abundance to become the only practical listening environment.
There is evidence of that appetite for separation. Deezer’s own research found that nearly half of music-streaming users would prefer AI-generated music to disappear from their feeds altogether. That does not establish that every listener will activate the setting, and it does not settle broader disputes over AI’s role in music. It does show why a clear opt-out tool could be more useful than a subtle disclosure.
Fraud is part of the stakes
The company’s case is not solely about aesthetics, authorship, or whether listeners like generated songs. It has also reported that up to 85% of streams involving AI-generated songs were fraudulent.
Streaming fraud refers to plays generated or manipulated in a way that seeks to game a platform’s systems, potentially including the systems that determine compensation or perceived popularity. In a catalogue filled with extremely high-volume uploads, bad actors may be tempted to treat music as a numerical lottery: upload at scale, generate attention by improper means, and hope some fraction of the catalogue produces revenue or visibility.
That reported fraud figure should be read carefully. It concerns streams of AI-generated songs, not a claim that every AI-made track or every creator using AI is fraudulent. Still, the number helps explain why an effective detection system is consequential beyond a listener’s personal taste. Better categorization can support listener controls, while also making it easier for a service to identify patterns that deserve enforcement scrutiny.
For human musicians, the practical concern is competition for limited attention. Discovery feeds, autoplay slots and playlist space are not infinite. If a service can distinguish generated material and let users exclude it, people who deliberately seek human-made work have a clearer route to finding it. Whether that changes the economics of streaming is a larger question, but it could change the composition of what a filtered listener is offered.
The detection question behind the toggle
A filter is only as useful as the system that decides what enters it. Deezer has already introduced an AI-music detection tool and has offered it to competing services. The new toggle is therefore an extension of existing classification technology rather than a promise to sort songs using a simple artist self-description.
No automated detection system is perfect. Deezer says its system carries a “very low” risk of a false positive. A false positive would occur if the tool classifies a human-made track as AI-generated and removes it for a listener who has enabled the filter. The reverse error—failing to identify an AI-generated track—would mean synthetic music could still slip through the setting.
Those limitations deserve to be kept in view because the available information does not claim flawless detection. A low reported false-positive risk is encouraging, but the useful standard for listeners will be how reliably the setting works across music styles, upload patterns and the many unfamiliar tracks encountered through recommendations. The planned feature is a strong expression of user choice, not a guarantee that every classification decision will be beyond dispute.
How the approach differs elsewhere
Other major music services have taken disclosure-oriented steps. YouTube Music recently introduced a label for AI-generated tracks, while Spotify uses similar labeling on AI artist pages. Those measures can help listeners understand what they are seeing, but they do not do the same job as an exclusion setting.
Deezer’s planned control puts the decision nearer to the user’s everyday listening routine. Someone who is curious about AI music can leave it available. Someone who prefers recordings and compositions they understand as human-made can opt out without having to perform constant detective work. The funny thing is that the feature is technically presented as a toggle, but culturally it functions more like a bouncer at the door of your Discover mix: clipboard out, synthetic backlog waiting outside.
The difference may shape expectations across streaming. Labels provide transparency. Filters provide agency. Neither approach alone resolves difficult questions about detection, fraud, artist attribution or the place of AI-assisted creation. But a listener-facing choice acknowledges that disclosure is not necessarily enough for people who want a different feed altogether.
What listeners should expect by year’s end
Deezer says the switch will arrive in its app by the end of the year. The information available does not detail the exact menu path, whether the preference will apply identically to every part of the service, or how the company will handle tracks that combine human work with AI tools. Those implementation details will matter, particularly because “AI-generated” can cover a wide range of creative processes.
What is clear is the stated aim: listeners will be able to remove AI-generated music from their Deezer experience. In an era when platforms have often asked users to adapt to rapidly expanding catalogues and opaque recommendation systems, that is a rare invitation to make the catalogue a little smaller on purpose.
The development also lands amid a broader music-business conversation about who gets discovered, who gets paid and what streaming services owe audiences in the way of transparency and control. For more on changes at the label side of that business, see this report on Jim Roppo’s move to Warner Records Group.
For now, Deezer’s promise is straightforward: by the end of the year, the people who want a human-only listening lane should be able to select one. Whether rival platforms follow with their own filters may be the next test of how much control streaming audiences can reasonably expect.
