Fake AI Tracks Are Appearing on Real Artists' Pages, and Royalties Follow
Scammers are putting AI-generated songs on the official pages of real artists and collecting the streaming payouts, according to an investigation first published by 404 Media on Sept. 17.
The money from those fake tracks goes to the uploader, not to the artist whose name and page carry them. That diversion of royalties is the purpose of the operation. Engadget
The method is cheap and uses ordinary, legal tools. An operator makes a track with a song generator such as Suno or Udio, makes cover art with an image generator, then uploads the package through an open distributor such as DistroKid, a service that places music on streaming platforms for a fee, while falsely naming an existing band. 404 Media tested the route end to end using the Brooklyn band Lathe of Heaven and confirmed it works. Engadget
Once accepted, the fake does not appear in a parody account or under a look-alike name. It shows up as a normal new release from the real artist on Spotify, Tidal, Apple Music, Amazon Music and other services. The payment follows the upload. For a listener browsing the catalog, there is no visible break, such as a change in the internal artist ID that links tracks to a page, to warn of fraud.
Spotify told 404 Media that profiles that are not actively managed can be especially at risk. That detail reflects how the system works. Distributors supply descriptive data, including the artist identifier, at the time of upload. The streaming services largely accept that claim unless an artist, label or rights holder objects through a Spotify for Artists claim, the dashboard where artists verify and manage their profiles, or a takedown request. Unclaimed profiles have no one watching to raise that objection.
Spotify's stated policy is that vocal impersonation in music is allowed only when the impersonated artist has given permission. The company laid out that policy in a September 2025 update. Spotify In practice, enforcement depends on detecting the fake and on someone with the rights to report it.
The problem has moved past isolated cases. The group Odette Child began tracking it in July and has logged more than 300 AI-generated songs impersonating acts including Taylor Swift, other well-known figures and deceased jazz musicians. The pattern points to two targets. High-traffic names where even a small share of misdirected streams pays, and less-visited catalogs where new additions go unnoticed. Engadget
404 Media published its findings under the title 'I Hijacked a Real Artist's Spotify with AI Music. It Was Disturbingly Easy'. 404 Media
This is not the first clash between open uploading and royalty payments. In November 2024, AI-generated albums that did not belong to them appeared on the Spotify pages of Standards, Health, Annie and Swans, in a case tied to distributor data errors. In July 2025, false tracks appeared on the pages of artists including Blaze Foley and Toto, which led singer Leith Ross to publicly criticize AI music after manipulated songs appeared under that name. The Verge Billboard
Earlier versions used human-made recordings. Billboard reported in 2022 that operators diverted royalties by uploading their own copies of viral hits, sometimes under the same artist name, and separately that uploaders chased playlist discovery with tracks falsely billed as featuring popular acts with no involvement. Billboard Spotify has also faced claims that it gamed its own royalty system by creating and promoting in-house 'fake' artists, claims it denied. The Verge
The broader context here is that streaming fixed distribution and left identity loosely defined. Frictionless uploading through aggregators let independent music reach a global audience, but it moved checking to the margins. Distributors compete on speed and price, not on verifying artist identity. Platforms compete on having a complete catalog, not on vetting each release in advance.
In my view, that design will need to change without closing the openness that made it useful. Stronger claiming of artist pages, a verifiable link between distributor accounts and confirmed artist profiles, and automated checks for unexpected releases on inactive pages are all solvable engineering tasks. They cost money and add friction, which is why they have trailed behind fraud. Generative audio has cut the cost of making convincing filler to near zero, so the old gap now grows faster.
Worth flagging for people who build these systems is where the response can work best. Spotting fakes by sound alone will keep falling behind generation. Checking identity and payment holds more promise. Who claimed this artist ID, from which distributor account, to which payment account, and has that account delivered for this artist before. Those answers are already in upload records. Using them firmly would not end AI music, nor should it. It would make impersonation unprofitable, which leaves space for legitimate uses to grow.
