Spotify Removes Over 50 Million Low-Quality Songs from Recommendation System
Spotify announced on October 20 that it has removed over 50 million songs identified as low-quality content (slop) or spam from its recommendation system this year.
These songs remain on the platform and can be found through active user searches, but will no longer appear through algorithmic or editorial recommendations.
Previously, the platform had completely removed over 75 million songs marked as spam due to serious or repeated violations.
The identification criteria include listener behaviors such as frequent skipping, ignoring, or actively indicating a desire not to listen, as well as a combination of upload patterns, metadata, and signals from third-party AI detection tools.
Spotify stated that after the adjustments, listener complaints about recommended content have significantly decreased, emphasizing the priority of presenting music from real human artists while continuing to update the system to address changing upload strategies.
These measures may reduce the occupation of the royalty pool by low-quality content and enhance user retention, with capital potentially flowing towards content review and recommendation optimization tools, while those relying on bulk-generated and inflated uploads may face pressure.
Source: Public Information
ABAB AI Insight
Starting in 2024, Spotify will implement a 1,000-play threshold to limit low-play content from earning royalties, with an AI protection plan to be announced in September 2025, including the removal of spam songs and recommendation filtering, and further expanding the identification and removal of low-quality content in 2026.
The platform categorizes content into recommendable and non-recommendable based on a combination of listener behavior and detection signals, with capital pathways focusing on concentrating distribution resources on highly interactive human-created content to maintain subscription experience and royalty distribution efficiency, rather than averaging distribution to all uploads.
Similar to video platforms downgrading low-quality or duplicate content, or app stores reviewing bulk-generated applications, the current streaming media landscape is at a stage where, following AI's reduction of production thresholds, supply has exploded, necessitating a reset of quality thresholds for distribution.
Essentially, this is a technological substitution, shifting from the distribution of scarce content to the filtering of excess content, with the mechanism being that the platform adjusts recommendation weights to determine the chances of being heard by strangers, resulting in lower creation costs while discovery and exposure costs relatively increase.
ABAB News · Cognitive Law
Supply surplus amplifies platform distribution rights
Listener behavior becomes a quality signal
When creation becomes cheaper, being heard becomes harder