More than half of the music uploaded to Deezer each day is now fully AI-generated, and up to 85% of the streams those tracks receive are fraudulent. At peak in June 2026 the platform was receiving about 90,000 AI tracks daily, up from 75,000 in April. Yet AI music accounts for only 1% to 3% of what people actually listen to, because the uploads are largely not there to be heard. Beatdapp estimates at least 10% of global streaming activity is fraudulent, worth roughly USD 2 billion in misallocated royalties a year.
TL;DR
- Beatdapp estimates at least 10% of global streaming activity is fraudulent (Beatdapp)
- That represents roughly USD 2 billion in misallocated royalties annually (Beatdapp)
- Deezer received about 90,000 AI-generated tracks per day at peak in June 2026 (Deezer)
- That exceeded 50% of all new uploads (Deezer)
- Up from about 75,000 per day and 44% of uploads in April 2026 (Deezer)
- Up to 85% of streams on fully AI tracks were fraudulent in 2025 (Deezer)
- Catalogue-wide fraud was about 8% of all streams in 2025 (Deezer)
- AI music accounts for only 1% to 3% of total streams (Deezer)
- Deezer detected and tagged 13.4 million AI tracks during 2025 (Deezer)
- Detection accuracy is 99.8% (Deezer)
- The tool misses roughly 2 in every 1,000 AI tracks (Deezer)
- False positives run below 1 in every 10,000 human tracks (Deezer)
- Beatdapp partners include Universal, Warner, Spotify, and SoundCloud (industry reporting)
1. The Scale of the Problem
Two independent estimates bracket the size of streaming fraud, and they are close enough to be useful. Beatdapp estimates at least 10% of global streaming activity is fraudulent, representing roughly USD 2 billion in misallocated royalties annually, while Deezer’s catalogue-wide figure for 2025 was about 8% of all streams.
The gap between 8% and 10% is smaller than the uncertainty in either number, so the honest summary is that roughly one stream in ten is fake. What makes this consequential rather than merely embarrassing is the royalty pool structure. Streaming services distribute a fixed revenue pool in proportion to stream share, which means every fraudulent stream does not create new money but takes it from legitimate artists. Fraud in this market is not a cost absorbed by the platform; it is a transfer from working musicians to whoever is running the scheme.
| Metric | Value | Source |
|---|---|---|
| Beatdapp estimate of fraudulent activity | at least 10% | Beatdapp |
| Estimated misallocated royalties | approx. USD 2 billion annually | Beatdapp |
| Deezer catalogue-wide fraud, 2025 | approx. 8% of all streams | Deezer |
| Range across the two estimates | approx. 8% to 10% | Derived |
| Who bears the cost | legitimate rights holders | Derived |
| Mechanism | fixed royalty pool split by stream share | Derived |
| Beatdapp monitoring basis | suspicious streams per quarter | Beatdapp |
| Beatdapp partners | Universal, Warner, Spotify, SoundCloud | Industry reporting |
Platform context sits in our music streaming statistics. Source: Streaming platforms back AI music labels as fraud actors exploit self-reporting gap.
2. AI Uploads Crossed the Halfway Mark
The upload figures moved faster in 2026 than almost any content metric on record. Deezer received about 90,000 fully AI-generated tracks per day at peak in June 2026, more than 50% of all new uploads, up from roughly 75,000 per day and 44% in April.
That is a 20% increase in daily volume and a six-point increase in share across roughly two months. Working backwards, total daily uploads sit somewhere around 170,000 to 180,000 tracks, which is itself a striking number for a single mid-sized service. The crossing of the 50% threshold matters symbolically but the trajectory matters more: a category that was 44% of uploads in April and above 50% in June is not stabilising. For scale, Deezer detected and tagged 13.4 million AI-generated tracks across the whole of 2025, an average near 36,700 per day, less than half the June 2026 rate.
| Metric | Value | Source |
|---|---|---|
| AI tracks per day, June 2026 peak | approx. 90,000 | Deezer |
| Share of new uploads, June 2026 | more than 50% | Deezer |
| AI tracks per day, April 2026 | approx. 75,000 | Deezer |
| Share of new uploads, April 2026 | 44% | Deezer |
| Increase in daily volume | approx. 20% | Derived from Deezer figures |
| Increase in share | approx. 6 points | Derived from Deezer figures |
| Implied total daily uploads | approx. 170,000 to 180,000 | Derived from Deezer figures |
| AI tracks detected during 2025 | 13.4 million | Deezer |
| Implied 2025 daily average | approx. 36,700 | Derived from Deezer figures |
Source: Deezer newsroom, AI music tops 50% of daily uploads.
3. AI Tracks Are Where the Fraud Concentrates
The connection between the two trends is the most important finding in this dataset. Up to 85% of streams on fully AI-generated tracks were identified as fraudulent and demonetised during 2025, against roughly 8% across the catalogue as a whole, a concentration of more than ten to one.
The mechanism is straightforward once stated. Stream-farming schemes need large quantities of tracks that no real listener would recognise, so that artificial play counts do not stand out against genuine demand. Producing that catalogue used to require either licensing filler or generating it laboriously. Generative music tools removed that constraint, making the marginal cost of a new track effectively zero. The result is not that AI music is inherently fraudulent but that fraud operations found in it a supply of exactly what they needed, and the 85% figure reflects the intent behind the uploads rather than a property of the audio.
| Metric | Value | Source |
|---|---|---|
| Fraudulent share of streams on AI tracks | up to 85% | Deezer |
| Catalogue-wide fraudulent share | approx. 8% | Deezer |
| Concentration ratio | more than 10 to 1 | Derived from Deezer figures |
| Action taken on detected fraud | demonetised | Deezer |
| Period measured | 2025 | Deezer |
| Underlying driver | near-zero marginal cost of track creation | Derived |
| What fraud schemes require | unrecognisable filler catalogue at volume | Derived |
| What the figure reflects | upload intent, not audio quality | Derived |
Source: Deezer says more than 50% of daily uploads are AI-generated.
4. Uploads Are Not Listening
The gap between how much AI music exists and how much anyone hears is enormous, and it is the most reassuring number here. AI-generated music accounts for only 1% to 3% of total streams despite exceeding 50% of daily uploads.
The reason is editorial rather than technical. Deezer excludes AI-detected content from algorithmic recommendations and editorial playlists, which removes the discovery paths that generate the overwhelming majority of streams on any platform. A track that cannot enter a playlist or a recommendation feed depends entirely on direct search, and nobody searches for a track they have never heard of. This is worth stating clearly because the upload statistic circulates far more widely than the listening statistic, and on its own it implies a listener experience that does not exist. The catalogue is being flooded; the listening surface is not.
| Metric | Value | Source |
|---|---|---|
| AI music share of total streams | 1% to 3% | Deezer |
| AI share of daily uploads | more than 50% | Deezer |
| Ratio between the two | roughly 20 to 1 | Derived from Deezer figures |
| Excluded from algorithmic recommendations | yes | Deezer |
| Excluded from editorial playlists | yes | Deezer |
| Remaining discovery path | direct search | Derived |
| Effect on listener experience | minimal so far | Derived |
| Which figure circulates more widely | the upload share | Derived |
Discovery context sits in our Spotify statistics. Source: Deezer newsroom, AI-generated tracks represent 44% of new uploaded music.
5. Detection Performance
The credibility of every figure above depends on how well the detection works, and Deezer publishes specifics. The tool reports 99.8% accuracy, missing roughly 2 in every 1,000 AI-generated tracks and producing fewer than 1 false positive per 10,000 genuine human-made tracks.
The asymmetry between those two error rates is deliberate and correct. A false positive wrongly demonetises a human musician, which is a serious harm to an individual; a false negative lets one AI track through, which is a marginal dilution. Tuning the threshold to make false positives five times rarer than false negatives reflects that asymmetry. At the June 2026 rate of 90,000 daily AI uploads, a 2-in-1,000 miss rate still means roughly 180 AI tracks slip through undetected each day, so even excellent accuracy leaves a meaningful absolute residue at this volume. The tool was deployed in January 2025.
| Metric | Value | Source |
|---|---|---|
| Detection accuracy | 99.8% | Deezer |
| Missed AI tracks | approx. 2 per 1,000 | Deezer |
| False positives on human tracks | fewer than 1 per 10,000 | Deezer |
| Ratio of the two error rates | approx. 5 to 1 | Derived from Deezer figures |
| Implied daily misses at peak volume | approx. 180 tracks | Derived from Deezer figures |
| Tool deployment date | January 2025 | Deezer |
| Tracks tagged in first year | 13.4 million | Deezer |
| Design priority | protecting human artists from false positives | Derived |
Source: Deezer says 44% of songs uploaded daily are AI-generated.
6. One Platform Is Supplying Most of the Evidence
The single largest weakness in this topic is not any individual number but where they all come from. Deezer is the only major streaming service publishing detailed AI upload and fraud figures, which means a mid-sized platform is supplying most of what the industry knows.
Two caveats follow. Deezer’s catalogue and its position in the market may not be representative, and a service with a smaller share of global listening could plausibly attract a different mix of upload behaviour than the largest platforms. More importantly, every figure here is self-reported by parties with a commercial stake: Deezer benefits reputationally from demonstrating rigorous detection, and Beatdapp sells fraud detection to the industry it is measuring. None of that makes the numbers wrong, and Deezer’s willingness to publish specific error rates is more transparency than any competitor offers. It does mean there is no independent audit of any claim in this roundup.
| Metric | Value | Source |
|---|---|---|
| Major services publishing detailed AI upload data | 1 | Derived |
| Independent audit of the figures | none available | Derived |
| Deezer’s commercial interest | reputational benefit from detection rigour | Derived |
| Beatdapp’s commercial interest | sells detection to the industry measured | Derived |
| Transparency offered by Deezer | published accuracy and error rates | Deezer |
| Comparable disclosure from competitors | not published | Derived |
| Representativeness of the catalogue | unverified | Derived |
| Confidence in direction of the trend | high | Derived |
Creator-economy context sits in our creator economy statistics, and subscriber dynamics in our streaming churn statistics. Source: Streaming fraud crackdown 2026.
Summary: Streaming Fraud by the Numbers
| Metric | Value | Source |
|---|---|---|
| Global streaming activity fraudulent | at least 10% | Beatdapp |
| Misallocated royalties annually | approx. USD 2 billion | Beatdapp |
| Deezer catalogue-wide fraud, 2025 | approx. 8% | Deezer |
| AI tracks per day, June 2026 | approx. 90,000 | Deezer |
| Share of daily uploads, June 2026 | more than 50% | Deezer |
| AI tracks per day, April 2026 | approx. 75,000 | Deezer |
| Share of daily uploads, April 2026 | 44% | Deezer |
| Implied total daily uploads | approx. 170,000 to 180,000 | Derived |
| AI tracks detected in 2025 | 13.4 million | Deezer |
| Fraudulent share of AI track streams | up to 85% | Deezer |
| Concentration versus catalogue | more than 10 to 1 | Derived |
| AI share of total streams | 1% to 3% | Deezer |
| Upload share to listening share ratio | roughly 20 to 1 | Derived |
| Detection accuracy | 99.8% | Deezer |
| Missed AI tracks | approx. 2 per 1,000 | Deezer |
| False positives | fewer than 1 per 10,000 | Deezer |
| Tool deployed | January 2025 | Deezer |
Methodology and Sources
- AI upload volumes, upload share, fraud rates on AI tracks, catalogue-wide fraud, listening share, and detection accuracy all come from Deezer’s newsroom disclosures (July 2026 release, April 2026 release).
- Independent coverage of the same disclosures was cross-checked against the primary releases (TechCrunch, July 2026, TechCrunch, April 2026).
- Beatdapp’s fraud estimates and partnership details come from industry reporting rather than a primary Beatdapp publication (TechTimes, Chartlex, Digital Applied).
- Data watch: every figure here is self-reported by a company with a commercial interest in the subject, and none has been independently audited. Deezer is the only major streaming service publishing detailed AI upload and fraud data, so a single mid-sized platform supplies most of the evidence, and its catalogue may not represent the market as a whole. The 85% fraud rate on AI tracks is described as an upper bound, and the 1% to 3% listening share is a range rather than a point estimate. The June 2026 upload figure is explicitly a peak rather than a monthly average, so it should not be compared directly against the April figure as if both were averages; the direction of the trend is nonetheless clear. Beatdapp’s global estimates are extrapolations from the platforms it monitors and are not a census. Rows marked as derived are arithmetic or direct inference from published figures.
- Last updated: August 2, 2026. We update this roundup quarterly, and the next major refresh is expected when Deezer publishes updated AI upload figures.