Concerns about artificial intelligence replacing human musicians reflect anxieties that have circled the industry for decades, from synthesizers in the 1980s onwards. A coalition of major music organisations recently proposed that streaming platforms label recordings as either AI-generated or AI-assisted. However, this initiative may miss a fundamental reality: the overwhelming majority of music available on streaming services receives virtually no audience engagement, with nearly 90% of tracks accumulating fewer than 1,000 plays in a year.

The proposal faces significant practical obstacles. Distinguishing recordings as AI-assisted becomes increasingly difficult when contemporary music production routinely incorporates machine learning at multiple stages—through mastering, pitch correction, sample processing and beyond—making clear categorisation nearly impossible. Additionally, detection technology would struggle against continuously advancing generative models and deliberate misrepresentation, ultimately producing an unreliable labelling system whilst forcing streaming platforms to absorb substantial compliance costs without addressing core industry concerns.