The music casino: How prediction markets are threatening Spotify’s No. 1

On the night of June 28, trader Caleb Davies, 47, noticed unusual activity on Kalshi, the U.S. prediction market platform where users can bet on almost any real-world outcome, including which song will top Spotify’s U.S. chart each day.

The following day, streams of two songs surged exclusively in the U.S. market — an indicator of possible manipulation. He alerted the platforms involved. “I knew instantly it was fraud,” he told EL PAÍS in a written response.

Spotify removed the streams for those two tracks, but overlooked a third: Earrings, a song by U.S. singer-songwriter Malcolm Todd released in 2024. The track’s streams had jumped 70% in a single day, pushing it to No. 1 on Spotify’s U.S. chart. When Spotify later removed more than 500,000 streams attributed to bots and recalculated the rankings, the bets had already been settled. Kalshi had given the song only a 2.5% chance of reaching the top spot the previous week, meaning anyone who correctly backed it increased their money twentyfold. Neither Todd nor his team were involved in the scheme.

Spotify directs roughly two-thirds of every dollar generated by music to rights holders and, according to its own figures, one million streams was worth an average of $11,000 in 2025 (up from $1,000 in 2014). Little by little, the business of faking those streams has become industrialized. In 2025, the French platform Deezer reported receiving nearly 90,000 songs generated entirely by artificial intelligence every day, accounting for more than half of all new music uploaded to the service, and estimated that as much as 85% of the streams of those tracks were fraudulent.

The International Confederation of Societies of Authors and Composers (CISAC) estimates that generative AI could cannibalize 24% of music creators’ income by 2028. One such case recently made its way to court: producer Michael Smith, indicted by federal prosecutors in New York in September 2024 in the country’s first criminal case involving AI-assisted streaming fraud, later pleaded guilty to collecting more than $8 million in royalties since 2017 through thousands of automated accounts and hundreds of thousands of AI-generated songs. The money came directly from the streaming royalty pool itself.

Although this is not a new practice, until now those who artificially inflated streaming numbers were seeking gains tied to the music itself: greater royalties for songs they owned rights to or the appearance of a hit that could trigger recommendation algorithms. With betting platforms, however, a new actor enters the picture: the charts become a financial asset on which a third party can speculate, creating an economic incentive that goes beyond the machinery of the music industry itself.

“We don’t see it as an entirely new phenomenon, but as a financial evolution of some streaming dynamics,” says Alberto Arenal, 42, head of innovation at AIE, Spain’s artists’ rights society that manages performers’ royalties and publishes Tempo&Stomp, a newsletter about the music industry. “Streaming rankings cease to be only indicators of consumer behavior and become an underlying asset that can be speculated on. In that sense, the phenomenon is new.”

For Davies, the imbalance is what makes the situation dangerous. “Markets create a huge incentive for people to manipulate the charts,” he replies. “Artists have the same incentive, but not for the same immediate reward a prediction market offers, and the risk is much greater for the artist: if they’re caught, they risk losing their royalties or being removed from the platform. If a trader is caught, they don’t risk any relationship with Spotify.”

BMAT, the Barcelona tech firm that produces official streaming charts for more than 30 countries, applies an anti-fraud filter and excludes songs with signs of manipulation before publishing each chart. Its commercial director, Alex Loscos Mira, 52, describes some of the ways such fraud is detected: “A song that shows a million new plays on Spotify in one week but has no growth on YouTube would be inorganic,” he says in a written response.

For Loscos Mira, it all boils down to a fraud-versus-profit equation: “As long as what a fraudster pays per fake user is less than what they earn in rights from that fake user’s plays, there will be an incentive to commit fraud.” He suggests a solution: “Some believe switching to an artist-centric model, where a user pays only the artists they listen to, could end or reduce that incentive.”

The problem is that no one is overseeing the intersection between music and betting. “These cases fall between two jurisdictions: financial regulators, who don’t look at music charts, and platforms, which don’t police an external betting market,” says Arenal.

Outside the music industry, prediction markets are already the subject of broader controversy. They have expanded rapidly in the United States, where people can bet on almost anything, from the Oscars to weather forecasts, and they are engaged in legal battles with state authorities that regard them as a form of disguised gambling.

In Spain, such platforms are outright illegal. In May 2026, Spain’s Directorate General for Gambling Regulation (DGOJ) opened sanctioning proceedings against Kalshi and Polymarket and ordered their websites to be blocked. According to Spain’s Ministry of Consumer Affairs, their classification as “unauthorized gambling” stems from the fact that the services they offer fall within the legal definition of gambling under Law 13/2011, but lack the mandatory license required to operate. The ministry added that betting on which song will reach No. 1 on a chart would fit the law’s definition of “peer-to-peer betting.”

Spotify declined to answer EL PAÍS’s questions and instead referred to statements it had already made to the Financial Times, which first reported the case. According to Spotify, streaming services face manipulation that is “constantly evolving,” and the company maintains that it has “best-in-class” detection systems and does not pay royalties on fraudulent streams. Spotify also requested that Kalshi and Polymarket remove its logo from those prediction markets.

However, manipulating metrics is not always about making money. In late June, Last.fm, the platform that tracks everything users listen to through a process known as scrobbling, cleaned up play counts generated by bots. Many fans of the group BTS, who for years had boosted their idols’ numbers through automated streaming, saw those figures collapse. In response, they launched a wave of negative reviews that drove the app’s rating below three stars. In 2025, Spotify had already removed hundreds of millions of streams from K-pop artists after determining that the plays were artificial.

The drive to elevate an artist predates prediction markets; these platforms simply add a cash prize to the equation. “Prediction markets are still an emerging, speculative space and do not replace consumption metrics,” says Arenal, who warns about their knock-on effects: “Charts and monthly listeners reflect taste, but they also shape it. They act as social proof and feed a positive or negative feedback loop. A false signal injected into that loop gets amplified.”

Héctor Fouce, 54, a professor at Madrid’s Complutense University, frames it as a deeper shift — the move “from a productive economy to a speculative economy.” Numbers, he argues, also confer legitimacy: “When your taste matches the mainstream, what remains are the numbers.”

In that way, the cycle continues, allowing those with the greatest resources to influence what is perceived as popular. Monthly listener counts and the No. 1 spot on the charts function as markers of legitimacy and help determine what people choose to listen to. And if those figures can be manipulated to win a bet, the question remains: how much of what we believe is genuinely popular was, in fact, artificially engineered?

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition