AI generates over 50% of music uploads on Deezer: most are fraudulent

50% of the songs uploaded to Deezer in June 2025 were generated by artificial intelligence, and the vast majority of them are fraudulent, according to the French platform. This figure marks a turning point: in January 2025, AI accounted for just 10% of uploads. In six months, the proportion has quintupled, and the phenomenon shows no signs of slowing down. This article analyzes the historical context of streaming fraud, the impact of generative AI on the music industry, the consequences for artists, platforms, and users, and potential regulatory and technological solutions.

What has happened?

Deezer began monitoring the origin of uploads in January 2025, when generative AI was already starting to gain popularity among creators and scammers. Since then, the percentage of AI-generated songs has grown exponentially, exceeding 50% in June. The company attributes this increase to bots and stream farms that use AI to create low-quality songs and generate fake plays, with the goal of collecting royalties. According to the platform's own data, in January 2025, one out of every 10 songs uploaded was AI-generated; by June, the ratio was one in two. This growth is not linear: it accelerated starting in March, when tools like Suno and Udio became popular, allowing anyone to generate music with just a prompt. Deezer detected that most of these songs have generic titles like "Relaxing Piano" or "Lo-Fi Beats," durations of less than 30 seconds, and poor audio quality, making them easy for their systems to identify, but the volume of uploads—millions per month—makes complete filtering difficult.

Why is it important?

Streaming fraud is not new, but AI has scaled it to unprecedented levels. Until now, platforms dealt with silent tracks, short loops, or songs by ghost artists; now, AI allows for the production of thousands of unique songs in minutes, each with different metadata to avoid detection. This dilutes the revenue of legitimate artists and overloads detection systems. Deezer estimates that these fraudulent songs divert millions of dollars a year in royalties. To put it in context, in 2023, before the generative AI boom, the International Federation of the Phonographic Industry (IFPI) estimated that streaming fraud cost the industry between $200 and $300 million annually. With AI, that figure could multiply. Furthermore, the problem is not exclusive to Deezer: Spotify and Apple Music face similar challenges, although they have not disclosed public figures. A 2024 report by the analytics firm MusicWatch already warned that 15% of songs on streaming platforms could be fraudulent, and AI has only worsened the situation.

Consequences for the industry

• Artists: Lower streaming revenue, as the pie is split among more songs, many of which are fake. An independent artist who previously received $0.004 per stream now sees their share reduced because royalties are distributed proportionally to total streams. Additionally, unfair competition from AI-generated songs that mimic popular styles can displace human creators in algorithmic playlists.
• Platforms: Higher moderation costs and the need for more sophisticated detection tools. Deezer has invested in an AI detector that analyzes spectral patterns and metadata, but the cost of keeping it updated against the evolution of generators is high. Other platforms like Spotify have begun using digital watermarks, but they are not yet mandatory.
• Users: Risk of encountering AI-generated content in playlists, damaging the discovery experience. A 2025 study by the University of Southampton showed that 40% of users cannot distinguish between AI-generated and human music, but 70% prefer to know the origin. A lack of transparency can erode trust in platforms.
• Regulators: Potential interventions to demand transparency in labeling AI-generated content. The European Union has already included AI-generated music in its AI Act, which will enter into force in 2026, requiring platforms to clearly label synthetic content. In the United States, the AI Music Transparency Act was introduced in Congress in May 2025, but it has not yet been passed.

What should readers know?

Deezer has implemented an AI-generated song detector, but the volume of uploads makes it difficult to filter all content. According to The Next Web, the company removed more than 10 million fraudulent songs in the first half of 2025, but the rate of uploads is such that some slip through. Other platforms like Spotify and Apple Music face similar problems, although they have not disclosed figures. The industry is looking for solutions such as digital watermarks (e.g., the C2PA standard) or blockchain to track authorship. However, these technologies are not yet scalable and require cooperation between platforms, artists, and labels. In the meantime, listeners can be wary of songs with generic titles, unusually short durations (less than 30 seconds), or poor audio quality, such as background noise or a lack of dynamics. They can also use tools like Deezer's detector, available to premium subscribers, to verify if a song is AI-generated.

"AI is democratizing music creation, but it is also being exploited to defraud the royalty system," notes the Deezer report.

The Deezer case is a warning: without controls, AI could flood platforms with synthetic content, eroding the trust of users and artists. The technology is not the problem, but rather its fraudulent use. The solution lies in combining automatic detection, clear policies, and cooperation between platforms. Furthermore, regulators must act quickly to establish transparency standards, such as mandatory labeling of AI-generated content. Otherwise, music streaming could become an ecosystem where the artificial outweighs the human, not because of quality, but because of quantity and deception.