Apple Music Playlist Playground: AI playlist generator that doesn't listen to prompts

Apple Music Playlist Playground: An AI Playlist Generator That Doesn't Listen to Requests

Apple Music has launched Playlist Playground — an AI-based playlist generator that is supposed to select music based on text prompts. The feature is in beta mode, and initial tests show it performs noticeably worse than competitors. The Verge editor asked the service to compile a playlist in the genre "atmospheric instrumental black metal for work."

The result was telling: three metal tracks with vocals, one field sound experiment, ambient electronics, and a piece of doom jazz. Not a single track that exactly matched the request. There was no mention of vocals in the description — Apple Music added it.

"Instrumental" meant without words, but the algorithm completely ignored this condition. For comparison, the same request was given to YouTube Music. Its AI made an error only on the fifth track — it returned one track with vocals, while the first four strictly matched the genre.

This is not perfect, but significantly better than what Apple offered. In YouTube Music's case, the error was an exception to the rule, whereas with Apple Music — it was the rule itself.

Playlist Playground appeared as part of the iOS 26.4 update and is available to Apple Music subscribers. Users enter a text prompt — mood, genre, activity, vibe — and receive an automatically generated playlist.

The idea is not new: similar features already work at Spotify, YouTube Music, and Amazon Music. Apple entered this niche later than competitors, and judging by first impressions, the algorithm has not yet reached the required level. The problem is not just with the specific black metal test case.

Users online share similar stories: you ask for instrumental — you get vocal, you ask for quiet background — you get loud rock. It seems the algorithm relies on broad genre labels and ignores the nuances of the request. "Instrumental," "atmospheric," "for work" — these are specific semantic constraints that an AI system must respect.

For now, Playlist Playground does not do this.

Over the past two years, Apple has actively positioned itself as a company betting on personalized AI. Apple Intelligence is one of the company's main marketing themes. But between promises and practical results, there is still a gap.

Music recommendations are one of the most challenging cases for AI: this requires not only genre classifiers but also understanding the semantics of the request, listening context, and subtle subcategories within a genre. A separate question is what counts as success in such systems. If the generator's task is to suggest something "close enough," the algorithm might consider black metal with vocals an acceptable answer to a general black metal request.

But a user who explicitly asks for instrumental makes that very word a priority. This is where AI recommendations still fall short of human curation: they hear the word "black metal," but they don't hear the word "without vocals."

The feature is still in beta, and Apple could theoretically refine it before a wide release. Nevertheless, the launch was awkward: a direct competitor handled the identical request significantly better, and users noticed. For a company that builds its image on precision and quality user experience, this is a useful signal.