Spotify has removed more than 75 million spam tracks from its catalogue and excluded another 50 million from recommendations this year. The figures reveal the scale of unwanted music flooding streaming platforms, but they also raise a question for independent musicians: how does an algorithm distinguish an artist experimenting with technology from someone trying to manipulate the system?
In an October 20 transparency update, Spotify said it had spent the past six months modifying its recommendation systems to limit what it calls AI-generated slop and spam. The company examined music that listeners repeatedly skipped, ignored or reported as unwanted, then combined those patterns with information about uploads and suspected fraudulent activity.
The two figures describe different actions. More than 50 million tracks were excluded from recommendations during 2026, while the 75 million figure concerns recordings removed entirely for serious or repeated rule violations. Spotify had already disclosed the latter total in September 2025, covering the preceding 12 months. It has not provided an updated cumulative removal figure.
Its detection systems examine metadata, listening behaviour, upload frequency and information supplied by third-party AI detection tools. They also look for duplicate recordings, artificially shortened tracks and networks designed to manipulate playlists or search results.
Spotify acknowledges that AI detection technology has limitations. Its latest report does not disclose how often legitimate music is incorrectly classified, how many exclusions are reversed or whether particular languages and genres are more likely to trigger its systems.
That lack of detail matters because music discovery increasingly depends on recommendations. A track can remain available on Spotify yet become considerably harder for unfamiliar listeners to encounter if it disappears from algorithmic playlists.
For independent artists without substantial promotional budgets, that distinction could affect their ability to build an audience. It is also particularly relevant to musicians using unconventional production methods, synthetic instrumentation or AI-assisted recording tools. There is no published evidence in Spotify's update that these artists are being systematically penalised, but the company has not released sufficient data to assess that possibility.
Spotify's own policy makes an important distinction between using AI creatively and misrepresenting an artist's identity. Its September 2025 policy statement said responsible AI use should not automatically result in lower rankings.
The platform now allows musicians to disclose AI-generated vocals, instrumentation and production contributions through song credits. Tens of thousands of these disclosures are submitted daily, according to Spotify.
It has also introduced AI Persona badges for artist profiles representing synthetic identities rather than real people. Thousands have been labelled. Their music is excluded from recommendations by default, although listeners can still seek it out or follow the profiles.
Another initiative, Verified by Spotify, now covers more than 300,000 artists, over 70% of them independent. The company says these verified performers account for more than 99% of the artists listeners search for.
The financial stakes are substantial. Since 2024, Spotify has required a track to accumulate at least 1,000 streams in the previous 12 months before qualifying for recording royalties. The company argues that the threshold reduces opportunities for mass-upload schemes to divert payments from legitimate musicians.
But the combination of royalty thresholds and recommendation filtering means smaller artists face two separate hurdles: getting their recordings heard and generating enough streams to earn royalties.
Spotify has described an appeal process for AI Persona labels applied by its review teams. Its October update offers less clarity about remedies for tracks excluded as spam.
The next test of its anti-spam campaign will therefore concern accuracy as much as volume. Removing millions of unwanted recordings is measurable. Establishing that legitimate musicians have not been caught in the same filters will require a different kind of transparency.
