Who actually gets flagged?
Spotify spent the last year cleaning house. In the twelve months to September 2025 it pulled more than 75 million tracks it judged to be spam, and in the same breath it switched on a "music spam filter" that tags uploaders gaming the system and quietly stops recommending them. On paper it hunts behaviour: mass uploads, duplicates under different artist names, artificially short tracks, metadata and SEO tricks. The trouble is the dragnet does not know who is human. A working producer who leans on a permissive distributor, or who uploads a flurry of edits and alternates, can trip the same wires as a bot farm, and the punishment is the same: no playlists, no algorithmic push, a catalogue that stops being found. Spotify insists it is "rolling the system out conservatively" so it does not penalise the wrong people, but the most exposed are the small, real artists with the least leverage to appeal.
Why do the AI bands slip through?
Because the one signal that would actually catch them, that a machine made this, is the one Spotify refuses to enforce. Its AI disclosure is opt-in: a beta that opened on 16 April 2026 lets artists declare AI in their vocals, lyrics or production through their distributor, and Spotify itself concedes that "the absence of a credit doesn't mean AI wasn't used." So an anonymous AI catalogue that drip-feeds polished, normal-length tracks and buys no fake streams sails straight past a filter built to catch crude fraud. The case study already happened. The Velvet Sundown, a band generated end to end with Suno, complete with AI promo photos and a backstory, climbed to roughly 1.4 million monthly listeners in the summer of 2025. It was journalists, not Spotify's moderation, who unmasked it.
The platform polices the honest and misses the anonymous.
How big is the problem Spotify will not measure?
This is not a fringe. Deezer, which actually publishes the number, says about 28% of the tracks uploaded to it every day, more than 50,000, are now fully AI-generated. Spotify has never disclosed the equivalent figure for its own platform, and falls back on the line that AI makes up "a really, really small percentage of streams." Maybe so. But streams are not the only thing at stake: every AI upload is another claim on the same royalty pool, another result wedged between a listener and a human artist, and another track the spam filter has to reason about. Refusing to count it is a choice.
What should an independent artist do?
Protect yourself from the filter before it protects you. Keep your releases on a reputable distributor, not a bulk-upload mill Spotify already watches; space out your uploads instead of dumping twenty tracks at once; fill in real metadata and credits; and if you use AI anywhere in the chain, disclose it, because a missing credit now reads as something to hide. None of this should be the artist's burden. But until Spotify is willing to name AI as confidently as it names spam, the safest place to stand is on the right side of a machine that cannot tell the difference.
