The term AI slop describes low-effort, mass-produced AI content uploaded at scale rather than music created with a clear artistic purpose. In streaming, the concern becomes more serious when volume is combined with fraud, imitation, or attempts to game recommendation and royalty systems.
AI uploads can scale far beyond human output
The source cites Deezer estimating that 18% of daily song uploads, around 20,000 tracks, were AI-generated at the time described.
That figure does not mean all AI music is low quality or fraudulent. The problem is that generation makes it possible to create and upload large quantities of music very quickly, so bad actors can scale spam more easily.
The article describes the criticism with the phrase musical margarine charging for butter, capturing the concern that listeners may be offered imitation or automated filler without being told what they are hearing.
Fraud and disclosure are separate issues
Deezer responded by developing systems to detect and label AI-generated tracks. The source also says fraudulent tracks could be removed from recommendations and denied royalties.
The fraud problem can be substantial. The Guardian reported Deezer saying that up to 70% of streams of AI-generated music on its service were fraudulent.
That is a different issue from using AI creatively. A musician can use generative tools without manipulating streams, while a fraudulent operation can exploit automation specifically to chase payouts.
Ownership remains difficult when provenance is unclear
The source also raises a harder question about ownership. Generative systems learn from large bodies of existing music, while a finished output may not provide a simple map showing which training examples influenced each moment.
That makes provenance difficult to explain, especially when a generated track resembles existing work.
Transparency can address only part of the problem. Labels can tell listeners that AI was used, but copyright and ownership still depend on how the work was created and what protected material may be present.
The useful response to AI slop is therefore narrower than rejecting AI music as a category. Streaming services need to detect manipulation and impersonation, while creators who use AI seriously still have an incentive to make work that is identifiable, deliberate, and worth listening to.
