I use Claude (Anthropic's AI assistant) integrated with Spotify to get personalized music recommendations and create playlists. The experience highlights a significant limitation in the current integration: all playlist creation is routed through Spotify's AI playlist generator, which produces unreliable results.
The problems with the current AI-based approach:
• Hits token limits and fails silently, resulting in playlists that don't appear in my library
• Substitutes incorrect tracks when it can't find exact matches, with no notification
• Generates generic results that don't reflect the specific tracks being requested
• Slow and resource-intensive compared to a simple database operation
The irony is that I get significantly better music recommendations from Claude than from Spotify's AI playlist feature. But when it comes to actually building the playlist, we're forced to route through a less capable AI layer that introduces errors and failures.
The fix is straightforward: a direct track-add API endpoint that accepts track IDs and adds them to a playlist. Licensing and content controls are already solved at the track metadata level and don't require AI interpretation. This would be faster, cheaper for Spotify to run, more reliable for users, and produce better results than the current AI middleman.
This isn't a request to remove AI playlist features — keep them for users who want them. This is a request for a reliable programmatic alternative that lets third party integrations like Claude deliver their full value to Spotify users.