Spotify‘s quest to curb the chaos AI-generated music has brought onto the platform has caused it to remove 50 million AI slop tracks from users’ recommendations, a new figure shared Tuesday as part of the company’s public insights into how it’s navigating the AI era.

The disclosure came in a post on its “For the Record” blog, where Spotify laid out the efforts it’s taken since it promised to push back against various forms of AI slop and spam that have proliferated across the streaming service. The 50 million tracks removed from recommendations were in addition to more than 75 million tracks removed from Spotify completely after being flagged as spam, though tracks that comprise the 50 million figure have not been removed from the platform.

“Earlier this year, we looked closely at the tracks listeners were skipping, ignoring, or telling us they didn’t want to hear,” the platform shared in the post. “A clear pattern emerged: these tracks were consistently ones identified as slop or spam. Over the past six months, we’ve made numerous updates to our recommendation system to remove slop and spam, and surface more music made by authentic, human artists. Our editors hold the same standard and don’t program content identified as slop or spam. Since we made these changes, we’ve seen a marked decrease in listener concerns about this content showing up in recommendations.”

The recap included progress reports on some of Spotify’s other efforts to combat AI content, including the ability to see which elements of a song were produced with AI within a track’s credits and the impact of its “AI Persona” labels. Spotify said “thousands” of profiles were now labeled as AI Personas, characters that are also excluded from recommendations unless a user follows their profile. It also said thousands of artists have opted into a program that allows them to screen which songs using their name appear on their profiles, an attempt to block the upswell of AI-modified variants of songs that invoke the original musician‘s names. AI-generated songs are also ten times as likely not to meet the 1,000-stream threshold to earn royalties, the company revealed.

“This work isn’t finished,” the company added. “AI moves fast, and bad actors adapt just as fast. But as AI makes music more abundant, we believe listeners will value human artistry even more. Our job is to make Spotify the place they find it.”

The efforts come as several music streaming services have tried to increase transparency about which songs scattered among their libraries have been produced by AI. YouTube Music last week joined Apple Music and Tidal in introducing labels for AI-powered tracks, while the latter has promised not to allow monetization for any AI-enabled song on its platform.