Spotify Removes 50 Million AI Slop Tracks From Recommendations
- AI-generated slop and spam don’t belong in your recommendations, and we’ve made changes to keep them out and protect the listening experience, Spotify said in an update on...
- Spotify stated that it analyzed user behavior over the prior months by looking closely at tracks listeners skipped, ignored, or actively rejected.
- To identify the unwanted audio, Spotify relied on listener engagement and feedback, upload patterns, metadata, and information gathered from multiple third-party artificial intelligence detection tools.
AI-generated slop and spam don’t belong in your recommendations, and we’ve made changes to keep them out and protect the listening experience, Spotify said in an update on Wednesday. The digital music streaming service announced that it has removed more than 50 million spam and artificial intelligence-generated tracks from its user recommendation systems, 1News reported.
Removing 50 Million Spam Tracks from Recommendation Systems
Spotify stated that it analyzed user behavior over the prior months by looking closely at tracks listeners skipped, ignored, or actively rejected. A clear pattern emerged showing that these rejected songs were consistently identified as spam or synthetic noise. Over the preceding six months, the platform implemented numerous updates to its recommendation algorithm to filter out those items and surface more music produced by authentic, human artists. Platform editors applied the same standard, declining to program content identified as slop or spam. Since introducing those changes, Spotify reported a marked decrease in listener complaints about unwanted tracks appearing in feeds.
Detecting Synthetic Uploads Through Engagement Patterns and Third-Party Tools
To identify the unwanted audio, Spotify relied on listener engagement and feedback, upload patterns, metadata, and information gathered from multiple third-party artificial intelligence detection tools. The platform previously removed more than 75 million spammy tracks last year for serious or repeated violations of its rules. This cleanup follows a broader industry debate over the rapid rise of generative software. Earlier in 2026, an artificial intelligence-generated song titled Celebrate Me climbed to the top of iTunes charts in the United States, the United Kingdom, France, Canada, and New Zealand after going viral on TikTok. Credited to an artist named IngaRose, the track was created on the American generative music software Suno using text prompts and amassed more than 7 million plays on Spotify.
Fears Over Unauthorized AI Training Data and Copyright Protection
Music creators and industry organizations have voiced strong concerns that existing technology does not sufficiently protect their work from unauthorized scraping and replication. Anthony Healey, division head of the music rights management organization APRA AMCOS New Zealand, told 1News that generative music software developers scrape existing catalogs without permission.
What we’re seeing at the moment in music is AI platforms that have taken essentially every song that’s ever been written, they’ve scraped it, they’ve copied it, they’ve trained their machine.
Anthony Healey