Your Favorite Artist Has a Fake on Spotify, and Nobody Can Stop It

There are fake versions of real artists on Spotify right now. Not covers. Not tributes. Not “sounds like” playlists. Tracks that clone an artist’s voice, style, and production aesthetic close enough to fool the algorithm — and sometimes close enough to fool listeners — sitting in search results next to the real thing, siphoning streams and royalties that should be going to actual musicians. Deezer says 75,000 AI-generated tracks are being uploaded to its platform every single day. That’s 44 percent of all new music uploaded. The platforms are losing this battle in real time, and the legal frameworks that were supposed to protect artists from exactly this scenario don’t yet exist at any meaningful scale.

The royalty math is the part that should bother you most. Streaming revenue is a fraction-of-a-cent-per-stream business for most artists. The only way it works at all is if the streams are real, the listens are genuine, and the algorithms surface the right music to the right people. AI cloning breaks all three conditions simultaneously. Fake tracks steal streams. Up to 85 percent of the streams those fake tracks generate are themselves fraudulent — bots and click farms amplifying the deception. And once fake content enters the algorithm, it competes for placement that would otherwise go to human-made music. A musician in North Carolina named Michael Smith figured this out before most people were paying attention: he built hundreds of thousands of AI-generated songs, ran them through bots, and defrauded streaming platforms of $10 million before getting caught. The tools to do this at scale are now accessible to anyone.

The platforms are trying. Deezer is detecting and tagging AI tracks, pulling them from algorithmic recommendations, and demonetizing fraudulent streams. Spotify removed more than 75 million “spammy tracks” in the past year and is testing “Artist Profile Protection” — a system that lets artists approve music releases before they appear on their official profiles. These are reasonable responses. They’re also insufficient for the scale of the problem. When 75,000 fake tracks are being uploaded daily to a single platform, detection and removal becomes a game of whack-a-mole that the platforms are structurally not equipped to win. Spotify’s removal of 75 million tracks sounds dramatic until you realize that number represents roughly one year of an automated system trying to keep up with a fire hose.

The legal picture is a patchwork. Tennessee passed the ELVIS Act in 2024, protecting vocal likeness from unauthorized AI cloning — the first state law specifically designed for this problem. California followed with its own law effective January 2025. At the federal level, the NO FAKES Act has been proposed to create nationwide protections for voice and likeness, but it’s still working through the process. Major labels — Universal, Sony, Warner — have filed lawsuits against AI companies for using copyrighted music to train their models, and some are pivoting toward licensing deals with conditions attached. None of this meaningfully addresses what’s happening on the platforms right now, today, where independent artists without label lawyers behind them have essentially no recourse when someone clones their voice and starts cashing in.

The independent artist problem is where this gets genuinely dark. Major label artists have legal departments, PR teams, and industry relationships that give them at least some leverage in this fight. Indie musicians have their Bandcamp account and a Twitter following. If someone builds a convincing AI version of a mid-sized indie artist and starts uploading tracks under a fake name, there’s currently no efficient mechanism for that artist to detect it, report it, or have it removed at a pace that outstrips the damage being done. By 2028, researchers estimate that AI substitution could put 24 percent of music creator revenue at risk. That’s not a speculative number. It’s what happens if the trajectory of the past two years continues without meaningful intervention.

The problem with framing this as a tech problem — something the platforms need to solve with better detection — is that it’s actually a market design problem. Streaming platforms are incentivized to maximize content volume, engagement metrics, and subscriber retention. AI-generated music, even fraudulent AI-generated music, contributes to all three until the fraud is detected. There’s no obvious mechanism by which Spotify or Deezer is financially motivated to solve this faster than they have to. The artists getting hurt don’t have leverage over the platforms the way major labels do. And the AI companies generating the tools that make this possible have, until recently, faced essentially no legal accountability. Something in that structure has to change — either through regulation, platform policy, or the kind of lawsuit outcomes that make the status quo expensive enough to abandon. Right now, none of those mechanisms are working fast enough.