AI Music Filter for Spotify
Spotify's algorithms keep pushing AI-generated tracks into your library. RealSongCheck finds them, removes them automatically, and skips them the moment they play.
/auth/callback. Copy your Client ID.Found a track that slipped through? Report it here. Every report gets checked against our detection signals. If it checks out, the track and its artist are added to the community blocklist for everyone.
We pulled a track you're certain was made by a real person? You can push back here. Every appeal gets read by a human. We update our database when we get it wrong.
At some point in 2023, Spotify's recommendation engine started pushing AI-generated tracks into regular people's libraries. Discover Weekly. Release Radar. Daily Mixes. Tracks churned out by tools like Boomy, Mubert, Suno, and Udio — no human behind them, no artistic intent, just content farms gaming the algorithm for streams.
The problem goes beyond taste. Every stream on an AI-generated track pays royalties to whoever uploaded it — money that would otherwise go to a working musician. Spotify has done some housekeeping, but the farms move faster than any moderation team can manage.
RealSongCheck is a small counter-measure. Connect your Spotify once, and from that point it quietly watches your library, removes what it finds, and skips anything that plays. No subscription required.
No single signal is definitive. We score tracks across several factors and only act when enough of them line up.
We only auto-remove tracks with a high confidence score. Medium and low risk ones get flagged for you to look at. We make mistakes — the is there for exactly that reason.
Every AI artist flagged by any RealSongCheck user feeds into a shared database. The moment an artist is confirmed, they're added to the global blocklist and every user's library is cleaned accordingly. New users benefit the moment they sign up — their first scan already runs against everything the community has found.
RealSongCheck works almost entirely from context signals — metadata, distributor patterns, release behaviour, and community reports. We do not do real-time audio analysis on individual tracks. Here's why that matters:
In theory, AI-generated audio has detectable acoustic fingerprints — characteristic artifacts from tools like Suno, Udio, or Mubert. In practice, detecting these requires access to the raw audio file (Spotify doesn't give third-party apps the raw audio stream), plus a trained audio classifier running on a server. That's technically possible but expensive. If this service grows, audio fingerprinting is on the roadmap. For now, the context-signal approach catches the vast majority of AI farm content because those operations have very consistent metadata patterns.
We don't add tracks. We don't sell data. We don't share anything beyond what Spotify's API requires to function. There are no ads and no investors. If RealSongCheck has been useful to you, covers the server bills.
Music royalties run on fractions of a cent per stream. One AI farm with a thousand fake tracks, each racking up a few hundred streams a week, pulls in real money. That money comes directly out of the pot that pays human musicians. RealSongCheck is one small way to push back.
Every feature works whether you pay or not. If you find it useful and want to help keep the servers running, you can contribute what you want — or nothing at all.
100% of contributions go toward server costs and the AI artist database. We don't take a cut.
These aren't vanity metrics. Each track removed represents a stream that won't go to an AI content farm — and therefore a fraction of a cent that stays in the royalty pool for working musicians.
That fraction matters at scale. Spotify pays roughly £0.003–£0.005 per stream. A single AI farm operating at modest volume — a thousand tracks getting a few hundred streams each per week — pulls in several hundred pounds a month. That money comes directly from the pro-rata pool that pays human artists.
We don't show artist names or track names here. This page exists to reflect the scale of the problem and the scale of the pushback — not to platform the content that caused it.
| Parameter | Type | Description |
|---|---|---|
| id | string | Spotify track ID (e.g. 4iV5W9uYEdYUVa79Axb7Rh) |
| Parameter | Type | Description |
|---|---|---|
| id | string | Spotify artist ID |
| Parameter | Type | Description |
|---|---|---|
| limit | int | Max results (default 20, max 100) |
| offset | int | Pagination offset |