Deezer is giving listeners an off switch for AI music. On October 14, 2026, TechCrunch reported that the streaming service is building a toggle that stops AI-generated tracks, albums and related artists from appearing in search results, with a launch expected before the end of the year.
It is a small product change with a large background: AI-generated tracks now make up over half of the music uploaded to Deezer each day. Below we cover what is confirmed, the numbers, how other services compare, and what a detection-based filter can and cannot promise.
What is confirmed
| Item | Detail | Source |
|---|---|---|
| Feature | Toggle that hides AI-generated tracks, albums and related artists in search | Deezer statement via TechCrunch |
| Timing | Before the end of 2026 | Deezer |
| Motivation | CEO Alexis Lanternier cites user demand and the explosion of AI music | Deezer |
| Demand data | About 45% of customers wanted a way to remove AI music, per a survey from last year | Deezer, via TechCrunch |
| Upload share | More than 50% of daily uploads were AI-generated in July, about 90,000 a day | Deezer, via TechCrunch |
| Earlier baseline | 10% of daily uploads when Deezer began reporting last year | TechCrunch |
| Not yet known | Accuracy, false-positive rate, whether it covers playlists and radio | Not announced |
Lanternier said the company is "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."
Uploads are not listening
The 50% figure describes what gets uploaded, not what people play. Cheap AI tools let a single person generate thousands of tracks, and many uploads are aimed at gaming streaming payouts. Deezer's earlier reporting and tools focused on fraud: it tags albums with fully AI-generated tracks, leaves them out of editorial playlists and algorithmic recommendations, and, per reports, filters fraudulent streams out of royalty payments. We discussed that scale problem in Spotify's labeling of AI artists and slop music.
So the toggle matters less for the average listener's day-to-day (they rarely encounter AI tracks by chance) and more as a trust feature: it lets people who care about human-made music choose to see only that.
From tags to a toggle: how Deezer got here
Deezer's stance has escalated in steps. In June 2025 it began tagging albums that contain fully AI-generated tracks, with a notice that some tracks "may have been created using artificial intelligence", framed at the time as a response to streaming fraud. Fully AI-generated tracks were kept out of editorial playlists and algorithmic recommendations, and fraudulent streams were filtered from royalty payments. In June 2026 the company released a free detector tool that lets people check playlists they built on other services. The October toggle is the first step that hands the decision to the listener rather than the platform.
That progression tracks the numbers. Deezer says AI tracks were 10% of daily uploads when it started reporting, and over 50% by July 2026. When a majority of new uploads come from automated pipelines, a recommendation tweak is no longer enough: the catalog itself is changing character. Offering a filter also manages a business risk. If listeners feel the service is filling with slop, they may leave for services with stricter policies, such as Bandcamp.
The economics behind the flood
Why are so many AI tracks uploaded? Streaming payouts are pro rata: a pool of subscription revenue is divided by plays. A very large catalog of cheap tracks, each earning a tiny amount, can add up if the tracks are streamed, including by bots. That incentive explains why Deezer pairs listener controls with fraud filtering. Genuine AI music made for listeners, meanwhile, is a different category, with tools such as Suno and Lyria producing songs people actually want to hear. A toggle treats both the same way, which is its main conceptual limit: it filters by how a track was made, not by whether it is good or whether it is fraud.
For rights holders, the stakes are different again. Labels are suing and licensing at the same time, as the Suno and Anthropic cases show. A visible AI flag on a platform strengthens the argument that the industry needs consent and credit systems, since it makes the provenance question concrete for ordinary users.
What artists can do
- Know the label policy. Check how your distributor and each platform treat AI-assisted work, and disclose where required.
- Keep stems and project files. If a track is wrongly flagged as AI, evidence of the production process is your best appeal.
- Ask about appeals before the launch. Deezer has not described a process for contesting a flag; artists and trade groups should press for one.
- Watch the definition. "Fully AI-generated" and "AI-assisted" are different, and the filter's scope will decide who is affected.
How the big services compare
According to TechCrunch's roundup:
| Service | Approach to AI music |
|---|---|
| Deezer | Detection tooling, tags, exclusion from recommendations, and now a filter toggle (coming) |
| Bandcamp | Complete ban on AI tracks |
| Tidal | Cut monetization for people uploading AI music |
| Spotify | Leaner stance: tags AI tracks and artists more prominently |
| Apple Music | Reported by Variety in August to be planning AI labels by year end |
Policy is also moving outside the platforms. Australia's chart body has set rules on AI music, covered in our post on Australia's AI music charts ban and ARIA rules. And the industry is splitting between litigation and licensing: see Suno v6 and the Warner and BMG settlement and Sony and Warner's lawsuit over music training. Generators keep improving, as in Google's Lyria 3.5 in Gemini, which makes detection a moving target.
How a filter like this works, and where it breaks
Detection systems typically look for statistical traces left by generative models in audio, or use known fingerprints from popular generators. A filter then hides items flagged above a threshold. Three limits follow:
- Arms race. As generators improve and new ones appear, detectors need retraining. A model that avoids known artifacts will slip through.
- Hybrid tracks. Human songs with AI vocals, AI stems or AI mastering sit in a gray zone. A binary toggle cannot capture it, and Deezer's tagging has focused on fully AI-generated tracks.
- Mistakes both ways. False negatives let AI tracks through; false positives hide real artists, a serious problem for small musicians who cannot appeal easily. Deezer has not published error rates for the toggle.
Questions worth asking when it ships: Is the setting global or per profile? Does it apply to playlists, radio, podcasts and recommendations or only search? Can artists contest a flag? How often is detection updated?
What it means for artists and listeners
For listeners, the toggle is a choice, and it is likely to be popular given Deezer's survey. For human artists, a visible human-only mode could make the human-made label a positioning advantage, a theme we explored in The AI Human Premium. For AI music creators, it is a risk: tracks flagged as AI may be hidden from users who opt in, narrowing the audience, though the music stays on the platform for those who do not use the filter.
For platforms, Deezer is making AI-versus-human a product axis, not only a fraud-control problem. If listeners respond, expect competitors to match with their own controls; Apple Music's planned labels point in that direction.
Questions people are asking
Does the filter delete AI music? No. It hides it for users who enable it. The reports describe it as removal from a user's own experience, not a ban.
Is it on by default? Not stated. The announcement calls it a toggle users can turn on.
Will it cover my existing playlists? Unknown. The reported scope is search results, tracks, albums and related artists.
What about AI-assisted human music? Deezer's tags have targeted fully AI-generated content, so AI-assisted tracks may not be flagged. Policy detail has not been published for the toggle.
Can I check other services? Reports say Deezer launched a free detector in June 2026 for checking playlists on other platforms. Treat results as indicative, not definitive.
What to watch before launch
Three details will tell us whether the toggle is a serious feature or a headline. First, whether it is a profile-level default that persists across devices, since a setting buried in a submenu changes little. Second, whether Deezer publishes accuracy figures and an appeals route, because hiding a real artist by mistake is costly and public. Third, whether the filter extends to Flow, the personalized stream, and to podcasts and audiobooks, where synthetic voices are also spreading. Independent testers will likely probe the detector with known AI tracks and with human tracks processed through AI tools; their results will say more than any press statement.
Bottom line
Deezer's toggle turns a content-moderation problem into a listener control. It is not live yet and its accuracy is unproven, but the numbers behind it, over half of daily uploads being AI-generated, explain the move. Watch for the launch details before year end: scope, defaults, error handling and whether competitors follow.
Details reflect TechCrunch's October 14, 2026 report on Deezer's announcement; the feature is not yet released.