AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Deezer will soon let you filter out AI-generated music

Collected Oct 14, 2026

Deezer said on Tuesday that it is developing a toggle that lets listeners filter AI-generated music out of the platform, and the company expects the feature to be available before the end of the year. When a user turns the toggle on, Deezer says AI-generated tracks, albums and related artists will no longer appear in search results. CEO Alexis Lanternier framed the move as a response to what the company describes as clear user demand for the ability to remove AI-generated content entirely, so listeners can focus on human creators if they choose.

The demand signal Deezer cites is a survey from last year in which around 45% of its customers said they wanted a way to strip AI music from their feeds. Lanternier also pointed to the sheer volume of AI music arriving on the platform as one of the catalysts for the toggle.

That volume has grown sharply. Deezer added AI-detection tooling last year and has published upload statistics since then. At the time the detection system launched, AI-generated tracks accounted for 10% of daily uploads. By July, the company said AI-generated tracks made up more than half of its daily uploads, averaging around 90,000 tracks per day.

Those two numbers together tell a story worth pausing on. If roughly 90,000 AI tracks land each day and they represent over half of everything uploaded, total daily uploads are on the order of 160,000 to 180,000 tracks. For a catalog team, that is not a curation problem in the traditional sense; it is a supply problem. Traditional streaming economics assumed uploads were expensive enough to act as a quality filter. Generative music tools have made the marginal cost of a release close to zero, so the filter has to move downstream, into detection and ranking.

Deezer's approach depends entirely on that detection layer working. The toggle only helps if the platform can reliably tell which tracks were made by AI, and that classification is the hard part. Detection signals are typically indirect: metadata, distribution patterns, audio forensics, and increasingly the presence of specific vocal or production artifacts. Each of these can be gamed, and each produces both false positives and false negatives. A false positive hides a human artist's work from a listener who asked to see only human work, which is a real harm to that artist, not just a product bug.

Deezer's peers have chosen different points on the same spectrum. Bandcamp has banned AI tracks outright, the most restrictive position. Tidal has kept AI music available but cut off monetization for people who upload it. Spotify has taken a lighter approach, tagging AI tracks and artists more prominently rather than removing them. In August, Variety reported that Apple Music was aiming to label AI tracks by the end of the year. Alongside platform-level tagging, some listeners rely on independent tracking systems to identify AI music.

The spread of these policies is instructive for anyone building recommendation or search surfaces. There is no industry consensus on whether AI music should be blocked, demonetized, or labeled, and the choice carries different costs. A ban is simple to state but hard to enforce at scale. Demonetization changes incentives without changing discovery. Labeling preserves choice but asks the listener to do the filtering. Deezer's toggle is closest to the labeling camp, except it goes a step further by letting the user remove the content from view rather than merely marking it.

A likely trade-off is discoverability. If a meaningful share of listeners switch the filter on, new AI artists lose access to exactly the audience most likely to be searching for them, while human artists get a cleaner search environment. Deezer has not said how many users it expects to enable the toggle, nor whether the filter will extend to algorithmic recommendations and editorial playlists in addition to search. Those are the places where filtering would matter most, and the announcement only covers search results.

Why it matters: developers and product teams building search, recommendation or moderation systems for audio are watching a live experiment in content classification at scale, with detection accuracy sitting directly on top of user trust. The 45% survey figure suggests a substantial share of listeners actively want filtering controls, which makes opt-out toggles like this a reasonable pattern to copy. Anyone shipping generative audio tools should also note that discovery surfaces are increasingly being asked to distinguish human from machine output, and that distinction may soon become a default expectation rather than an optional feature.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Deezer said that the new toggle will be rolled out by the end of the year.