Fake streams take 60% of AI-tagged plays at Qobuz, and a metric goes with them

What Qobuz announced, and the numbers under it#

The fox counts footprints, not shouting. On 24 September Qobuz announced a tag that alerts listeners when a release has been identified as AI-generated, and Music Business Worldwide set out the detail on 28 September. Tracks identified as AI-generated are 0.38% of all streams on Qobuz. Of the streams on those tracks, 60% are “currently deemed fraudulent by its anti-fraud systems”. And Qobuz has removed more than a third of the AI albums with no listening activity from its search engine, covering four million tracks.

The mechanics matter to anyone who markets music. The tag appears at release level. Content identified as AI-generated is not promoted in the areas of the app that rely on human editorial selection, including Our Selections, New Releases, Qobuzissimes and Albums of the Week, and it is excluded from algorithmic recommendations, according to the Qobuz AI Charter. Where fraud is detected, the streams are excluded from reporting and royalty payments, and the content is removed from the catalogue.

By our arithmetic, 60% of 0.38% is roughly 0.23% of all Qobuz streams. That is small, and it is exactly why the tag is about trust rather than volume: a quiet quarry can still spoil a pond. The identification system was built in-house, and the charter concedes that “no reliable solution exists today to identify all AI content with certainty, especially when partial human intervention occurred.” Hold that sentence in mind, because we return to it in section four. The direction is not new: Qobuz’s 26 February 2026 press release promised tags “in the coming months”, and September’s tag is that promise arriving.

Qobuz is not alone. The Deezer figures charted above come from its 29 January and 21 July 2026 announcements, the second of which put AI-generated tracks at nearly 90,000 a day and above half of new uploads at the June peak. The same company says AI music is between 1% and 3% of its streams. A great many uploads, very little listening: that gap is the whole story of fake streams.

Why fake streams are the point, not the side effect#

Deezer says the quiet part out loud. In its January 2026 announcement it wrote that “Generating fake streams continue to be the main purpose for uploading AI-generated music”, and that while only a small share of total streams comes from AI tracks, up to 85% of those streams were fraudulent in 2025, depending on the month. By comparison, streaming fraud across the entire catalogue accounted for only 8% of all streams that year.

Set the two Deezer numbers side by side and the incentive is visible. AI tracks are a sliver of listening and a large share of fraud, because the upload cost is close to zero and the payout, if nobody looks, is not. The royalty pool is shared by streams, so a stream that is not real still takes a slice. Spotify’s own artist page puts it in one line: artificial streams “dilute the royalty pool, shifting revenue from legitimate artists to bad actors”.

Spotify bot streams and the service that sells them#

The supply side is not mysterious. Spotify warns artists that services promising streams or playlist placement for a fee are scams, and that many of them use bots to stream a song on repeat. Spotify bot streams are the raw material of those services, and the same page says Spotify strictly prohibits using any third-party service that promises streams or playlist placement in exchange for money. If the promise sounds too good to be true, the page adds, it probably is.

Fake streams Spotify will not pay for#

The penalties are spelled out. When Spotify detects artificial streams, they do not earn royalties, they do not count toward public stream numbers or charts, and they do not positively influence recommendation algorithms. Depending on severity, the song can be removed from playlists, the distributor can warn, fine or suspend the account, and the track can be removed altogether. Since April 2024 Spotify also charges labels and distributors per track when flagrant artificial streaming is detected on their content, as the Spotify for Artists artificial streaming page explains.

What it means for music marketing: three discounts#

Marketing runs on numbers that look like proof, so the practical effect of tagging and exclusion is a set of discounts. The first is on your own numbers. Streams a platform judges artificial are excluded from public counts and charts on Spotify and from reporting and royalties on Qobuz, so a pitch deck built on inflated plays is now built on sand that the platform itself is sifting.

The second is on discovery lanes. On Qobuz, AI-tagged releases are kept out of the editorial spaces and the algorithm. On Deezer, AI-generated music is excluded from algorithmic recommendations. On Spotify, the AI Persona badge announced on 11 August arrives on profiles from mid-September, and by default those artists are kept out of editorial and algorithmic recommendations unless a listener follows them. A human artist with a machine-shaped catalogue could find a door closed.

The third discount is on vendors. Anyone who sells guaranteed plays, guaranteed playlist slots or guaranteed algorithmic pushes is now selling into a market where every major platform is publishing its exclusion rules. That does not make all music promotion suspect. It makes the burden of proof heavier, and it makes a written method a competitive advantage.

Early reaction from independent artists has been sharp, and fair. One posted within hours of the news:

The instinct to pay attention is the right one, and the answer is method, not panic. Our notes on the DistroKid AI lawsuit and licensed fan creativity show how quickly the rules around AI and artist marketing are moving, and our music industry marketing work starts from the same premise: earn the audience, then measure it honestly.

The wrong-tag risk and the upload fee argument#

Return to the sentence in Qobuz’s charter that admits detection is imperfect. It matters because a wrong tag costs a human artist their editorial lane. Qobuz says artists who believe an AI badge was assigned wrongly can contact their distributor, which handles those requests, and that anyone can report content they suspect is AI-generated but unlabelled through a form. It is also integrating the DDEX field for AI-generated content, so tracks can be identified when rightsholders declare them, with deployment under way across labels, distributors and aggregators.

That is the practical cue for a marketing team. Declaration through your distributor is becoming the clean route, and a missing or wrong declaration is a risk you can manage. Keep the paper trail: session files, stems, credits and any tool used to generate a part. If a badge lands wrongly, the evidence is what moves the appeal along.

The other live argument is about cost. On 28 September Music Ally reported that the independent-music think-tank ORCA wants an upload fee. Its executive director Patrick Clifton wrote that “It is time for upload fees to be instituted across the digital supply chain to disincentivise bad actors, as current approaches to the problem have not worked.” The proposal is a per-track fee for new recordings, modelled to find the optimal amount, with streaming services adding the fees to their royalty pools. Clifton acknowledged it would need to be “carefully calibrated” to avoid unintended demonetisation of legitimate artists and labels.

Whatever you think of the fee, notice its shape. It tries to raise the price of the exact behaviour that generates fake streams, which is high-volume uploading. Distributors are already moving in a similar direction: on 14 September IFPI launched the Streaming Integrity Initiative, backed by Sony, Universal, Warner, Merlin and a long list of distributors including CD Baby and Symphonic, with five commitments: verify rights and identities, vet content, act on suspected fraud, share intelligence and measure the systems.

How to spot fake streams in your own data#

The unglamorous work is looking at your own dashboard with a sceptic’s eye. Spotify’s artist page lists what abnormal activity may look like: a sudden spike in streams for no apparent reason followed by a drop-off; an unexplained spike from a location where your tracks have not been active; the majority of streams from unexplained or surprising sources, such as “Other”; and a sudden, short-lived jump in follower growth. Any of those, the page says, is a reason to tell your label or distributor and to share the details of any service you hired.

Two of those signals are marketing tells as much as fraud tells. A campaign that produces a spike and then nothing is either a fraudulent service or a campaign with no follow-through, and in both cases the fix begins by asking what the vendor actually did. That is why the audit below starts with the vendor, not the data.

Those five steps are unglamorous, and they are the cheapest reputation insurance in music. If you want a second pair of eyes on a campaign plan, a vendor list or a set of dashboards, our content marketing services and brand strategy work covers the honest end of artist growth, and our paid social team can show what a real, measurable fan campaign costs.

Frequently asked questions#