Spotify's War on "AI Slop"
If you've seen headlines this month claiming Spotify deleted 75 million AI songs, the real story is both bigger and more nuanced than that number suggests. Spotify isn't banning AI music, it's running one of the most aggressive anti-fraud campaigns streaming has ever seen, and the target is spam, not creativity.
The Headline
On July 23, 2026, Spotify confirmed to the Los Angeles Times and Rolling Stone that it has removed more than 75 million AI-generated tracks over the trailing 12 months, a volume roughly 1.5 times the size of Apple Music's entire catalog, averaging around 205,000 removals a day. This isn't a single dramatic purge, it's the disclosed output of a classifier that's apparently been running quietly at that volume for roughly a year, only now being talked about publicly as scrutiny of AI music intensifies.
What's Actually Getting Pulled
Despite the framing in some coverage, Spotify has been explicit and consistent: AI-assisted music is still welcome. What gets removed is content that behaves like spam or fraud, regardless of what tool made it. The targeted behaviors include mass uploads, duplicate tracks, SEO manipulation, artificially shortened songs designed to farm the 30-second royalty threshold, and outright impersonation. The economic motive is straightforward. Spotify pays royalties from a shared pool based on stream share, and every fraudulent stream from a bot-farmed AI track dilutes the pool that legitimate artists are paid from.
Spotify's enforcement framework rests on three pillars:
Impersonation enforcement. Unauthorized AI voice clones of real, recognizable artists are barred outright, and vocal impersonation is only allowed with the impersonated artist's consent. Spotify is also working with distributors to stop fraudulent delivery of AI-generated tracks to other artists' profiles.
A dedicated spam filter. Rather than removing everything automatically, Spotify tags tracks matching known abuse patterns, including thin distributor networks, fake metadata, sub-30-second runtimes, inflated stream counts, and cuts them off from algorithmic recommendation. If the abuse is blatant enough, the track comes down entirely.
AI disclosure via DDEX. Working with the nonprofit metadata standards body DDEX, Spotify built a credits-based disclosure system letting artists voluntarily specify whether AI contributed to vocals, instrumentation, or production. The AI Credits feature went into public beta on April 16, 2026.
Not everyone's convinced the system is precise. Spotify still has no consumer-facing button to filter out AI music altogether, relying instead on the voluntary disclosure system, which by definition can't catch artists who don't disclose. Community-run tracking projects have identified more than 4,700 accounts suspected of being AI-generated acts that continue operating on the platform like any other artist because they haven't triggered fraud or impersonation flags.
There's also a real false-positive problem. Because the spam filter reads technical signals, inaudible AI watermarks, off-spec mastering, or audio fingerprints that collide with existing tracks, rather than intent, some legitimate musicians who use AI tools as part of a genuine creative process report getting caught in the same net as spam farms.
And the policy has had real-world consequences beyond Spotify's own platform: in January 2026, the Swedish chart authority IFPI Sweden pulled the AI-assisted folk-pop hit Jag Vet, Du Är Inte Min by an artist called Jacub from the country's official singles chart after its AI origins came to light, even though the track remained live and chart-topping on Spotify's own in-app charts.
Not Just Spotify
Spotify's disclosure landed in the middle of a broader industry reckoning. Deezer has taken a stricter line, auto-detecting and tagging fully AI-generated uploads, which by its own count made up roughly 44% of daily uploads (about 75,000 songs a day) by April 2026, and demonetizing them rather than removing them outright. TIDAL announced that starting July 15 it would label fully AI-generated tracks and strip them of royalty eligibility, while still allowing them to remain on the platform.
Regulation is also catching up: the EU AI Act's Article 50 transparency requirements, mandating machine-readable labeling of AI-generated audio and deepfake disclosure, become legally enforceable on August 2, 2026, adding a legal backstop to what has so far been a voluntary, platform-by-platform patchwork of disclosure rules.
The Bottom Line
Spotify's 75-million-track figure makes for a dramatic headline, but the underlying policy is narrower and more deliberate than AI purge implies: it's a fraud and impersonation crackdown that happens to be sweeping up a lot of AI-generated content because that's where mass-produced spam is currently coming from. Genuine AI-assisted music, that is properly rights-cleared, honestly disclosed, and not designed to game the royalty system, remains explicitly welcome on the platform. The harder, still-unresolved question is whether voluntary disclosure and behavioral filtering can keep pace with a flood of AI content that may already account for close to half of new uploads across the wider streaming ecosystem.
Read more:
ABC - BTN Newsbreak 17/07/2026
Comments