AI-generated tracks have flooded music streaming platforms, but new data suggests many aren’t attracting real listeners. The growing problem raises questions about streaming fraud, artist royalties and what music popularity actually means.

Imagine releasing a song, uploading it to a streaming platform and watching the plays accumulate. The numbers suggest people are listening. There may even be enough activity to generate royalty payments. But what happens when the song wasn’t created by a human musician, and many of the supposed listeners aren’t human either?

That is the problem confronting the music streaming industry. In July, French streaming company Deezer reported that fully AI-generated songs had surpassed 50% of its daily music uploads on peak days in June. The platform was receiving approximately 90,000 AI-generated tracks each day, an extraordinary volume of new material entering a business already overflowing with music.

But the listening numbers told a different story. Fully AI-generated songs accounted for just 1% to 3% of total streams on Deezer. And in 2025, the company identified as much as 85% of streams involving those tracks as fraudulent. The result is an increasingly strange commercial environment in which music can be generated automatically, uploaded at enormous scale and played through manipulated listening activity without necessarily attracting a meaningful human audience.

The problem isn’t simply that computers can now create songs. It’s that the economics of streaming may reward activity that looks like popularity, even when that popularity has been manufactured.

Millions of Songs, Very Few Listeners

Deezer’s findings expose the difference between the amount of music entering a platform and the amount people actually want to hear.

In April, the company reported receiving nearly 75,000 AI-generated tracks daily, representing approximately 44% of new uploads. By June, that number had climbed to roughly 90,000, with synthetic music exceeding half of daily submissions on peak days.

That volume is possible because generative music platforms can produce finished recordings with relatively little time, equipment or technical knowledge. Instead of assembling musicians, booking studio time and recording individual performances, users can generate tracks through software and distribute them through services that deliver music to streaming platforms.

There are legitimate creative uses for that technology. Some artists experiment with synthetic sounds, use software to develop musical ideas or create recordings that would otherwise require expensive production resources. Fully generated music can also appeal to listeners interested in new forms of digital creativity.

But creating music and building an audience are different accomplishments. Deezer’s data suggests that the extraordinary supply of synthetic recordings has not translated into comparable listening demand.

The company’s findings apply specifically to Deezer, not the entire global streaming industry. They also concern fully AI-generated recordings, rather than every song involving some form of computer-assisted production.

Still, the figures demonstrate how rapidly the economics of digital distribution can change when the cost of producing new material approaches zero.

The Money Is in the Streams

Streaming services have made listening activity a central measure of commercial success. Every eligible stream contributes to the calculations used to distribute royalties, meaning recorded plays have financial value even when no individual listener purchases a song outright.

That creates an incentive to manipulate the numbers.

A fraudulent operator can upload large quantities of music, then use automated accounts or other artificial listening methods to generate plays. If those streams are counted as legitimate, the operator may receive royalty payments without attracting genuine listeners.

The combination of inexpensive AI-generated recordings and automated playback makes that model easier to attempt at scale. Instead of paying to produce hundreds of distinct recordings, an operator can generate them rapidly and attempt to monetize artificial activity across a large catalog.

Deezer says its systems identified up to 85% of streams involving fully AI-generated tracks as fraudulent in 2025. By comparison, approximately 8% of streams across its entire catalog were identified as fraudulent that year.

Those figures do not mean 85% of AI-generated songs were fraudulent or that everyone using AI music tools is participating in a scheme. They describe the share of streaming activity Deezer classified as manipulated, not the intentions of every creator uploading synthetic music.

The distinction is important because the financial problem is not simply the existence of computer-generated recordings. It is the attempt to convert artificial listening activity into money.

What Happens to Working Musicians?

For independent artists, earning a meaningful income from streaming is already difficult. Musicians must compete for attention, invest in recording and promotion, and develop audiences large enough to sustain their work.

Fraudulent streaming introduces another potential competitor for the money available within the system.

Many streaming services use royalty arrangements in which eligible listening activity influences how revenue is distributed. When fraudulent streams are improperly counted, they can divert funds that would otherwise support legitimate rights holders.

An artist who spends months producing an album may therefore be competing financially with an operator who generates recordings and manipulates playback at industrial scale.

However, the extent of that harm depends on whether platforms detect and exclude fraudulent activity before royalties are distributed.

Deezer says it removes identified fraudulent streams from royalty calculations, preventing those plays from taking revenue away from legitimate artists. The company maintains that its safeguards have kept the current impact on human musicians’ royalties minimal within its own service.

That is an important qualification. The company’s figures establish the scale of attempted manipulation on Deezer, but they do not prove that musicians across the industry have lost corresponding amounts of royalty income.

The broader concern is whether other streaming platforms can identify and prevent comparable activity before money changes hands.

Deezer Is Beginning to Remove the Music

Deezer has responded by attempting to separate genuine listener demand from automated activity.

The platform began deploying its AI music detection system in January 2025. It identifies technical characteristics associated with music-generation software and uses those signals to flag fully synthetic recordings.

Detected tracks are excluded from algorithmic recommendations and editorial playlists, reducing the likelihood that they will be automatically promoted to listeners.

In July, Deezer also announced plans to remove AI-generated tracks associated with streaming fraud and recordings that had not received a stream for at least six months.

The company has since expanded its response. In October, Deezer introduced an optional AI music filter allowing subscribers to exclude fully AI-generated tracks, albums and associated artists from search results.

According to the company’s latest figures, more than 42 million AI-generated tracks were detected and excluded from its recommendation systems between January 2025 and August 2026.

The new filter does not ban all synthetic music. Listeners who want access to AI-generated recordings can continue finding them, while those who prefer human-created music have another way to control their listening experience.

The approach also raises questions about how accurately technology can distinguish between fully synthetic songs and recordings made by musicians using AI-assisted production tools. Deezer says its system is designed to focus on fully generated tracks rather than ordinary human recordings incorporating limited AI assistance.

Are Listeners Even Interested?

The abundance of AI-generated music might suggest that listeners are embracing a new era of synthetic entertainment. Deezer’s research offers a more complicated picture.

In November 2025, the company commissioned an international survey with Ipsos involving 9,000 people across eight countries. Participants were asked about their attitudes toward AI-generated music and whether they wanted clearer information about how recordings were produced.

According to the survey, 97% of participants failed to consistently distinguish fully AI-generated music from human-made recordings in a blind listening test. At the same time, 80% agreed that completely AI-generated music should be clearly labeled.

Among streaming users surveyed, 73% wanted to know when a service was recommending fully AI-generated music, while 52% believed synthetic songs should not appear alongside human-made recordings in the main music charts.

The results suggest that listeners may struggle to identify synthetic recordings by sound alone while still caring about their origins.

That creates a new responsibility for streaming services. If listeners cannot reliably identify how music was produced, platforms may need to provide that information rather than expecting audiences to determine it themselves.

It also complicates the idea that technological advances automatically create consumer demand. The ability to produce enormous quantities of music does not mean audiences have asked for that music or intend to listen to it.

Music Popularity Is Becoming Harder to Trust

The streaming fraud problem connects to another developing issue in music promotion: the growing influence of paid recommendations disguised as organic enthusiasm.

As NPR reported in July, some music influencers receive payments from record labels and marketing agencies to promote songs without disclosing those arrangements to followers.

The two practices are different. Paid influencer marketing involves commercial promotion that may or may not be properly disclosed, while fraudulent streaming involves manipulating listening activity. But both reveal how easily the appearance of music popularity can become disconnected from authentic audience enthusiasm.

One can influence what people believe is worth hearing. The other can inflate the numbers suggesting that people are already listening.

Neither practice means every successful song is artificially popular. Music can still spread through genuine recommendations, organic discovery and listeners sharing songs they enjoy.

But as promotional campaigns and automated activity become more sophisticated, streams and online buzz increasingly require context. A large play count does not necessarily demonstrate broad public enthusiasm if some of those plays were generated through manipulation.

For artists, that raises questions about how success should be measured. For listeners, it challenges the assumption that popular music is always popular because people independently chose to hear it.

The Industry Has to Decide What Counts as a Listener

The recorded music business has survived major technological changes before. Radio, television, digital downloads and streaming each transformed how audiences discovered music and how artists earned money.

Generative AI introduces another challenge because it can affect both sides of the transaction. The music can be created automatically, and the listening activity can be artificially manufactured.

That combination threatens the reliability of the metrics used to measure demand and distribute revenue.

Deezer’s figures suggest that, at least on its platform, the volume of synthetic music has grown far beyond the audience interested in listening to it. The company’s decision to detect, label, filter and remove certain recordings represents an attempt to protect its catalog and payment systems from manipulation.

Other companies may adopt different approaches, and there is still a legitimate debate about the role AI-generated music should play in the creative economy. But there should be less disagreement about whether automated listening schemes deserve to receive money intended for genuine music consumption.

The question facing the streaming industry is no longer simply whether computers can create convincing songs. They can, and increasingly they do.

The more important question is whether the business can distinguish real musical demand from an automated system creating songs, generating plays and attempting to collect royalties without an audience ever showing up.