According to The Verge, Spotify is launching a beta feature called Prompted Playlists this Thursday, initially in New Zealand. The feature lets users type a text prompt—with any level of detail—to generate a custom playlist. Spotify’s AI then curates that playlist based on the user’s instructions and their personal listening history. Users can also set these playlists to regularly refresh with new songs based on the same prompt. This effectively creates a user-directed version of the platform’s popular Discover Weekly playlist. The move is part of a broader push for user control, mirroring a similar update to Spotify’s AI DJ in May that added voice prompts.
The Algorithm Flip
Here’s the thing: this is a pretty fundamental shift. For years, the promise of algorithmic discovery has been passive. You listen, it learns, it suggests. But the control has always been opaque, buried in likes, skips, and mysterious “taste profiles.” Prompted Playlists flips that script entirely. Now, you’re not just feeding the machine indirectly; you’re giving it direct orders. Want a playlist for “ambient synthwave that sounds like driving through a rainy Tokyo at 2 AM”? You can literally type that. It’s a move from implicit to explicit intent. And that’s a big deal because it acknowledges a simple truth: sometimes, we know exactly what we want, we just can’t articulate it to a traditional search bar.
Spotify’s Bigger AI Game
This isn’t a one-off experiment. Look at the pattern. Earlier this year, they added light genre controls to Discover Weekly. In May, they let you give voice prompts to the AI DJ. Now this. Spotify is methodically building an interface layer of conversational AI over its entire catalog and recommendation engine. The goal seems clear: reduce friction to finding the perfect audio for any moment. But there‘s a business angle, too. The more you engage with these AI tools, the more data you provide, and the better the core service gets at keeping you subscribed. It also subtly positions Spotify against platforms like Instagram, which is making similar “tell the algorithm what you want” moves. They’re all racing to solve the same problem: algorithmic fatigue.
The Real Test Will Be The Prompts
So, will it work? Well, that depends entirely on the AI’s understanding. The cool demo is the hyper-specific, cinematic prompt. But the everyday use might be “chill 90s rock” or “focus music for coding.” The risk is that it feels like a fancy search bar that occasionally misunderstands you. If you type “sad songs” and it gives you radio-friendly pop ballads when you wanted indie folk, you’ll probably just go back to your trusted playlists. The magic needs to be in the curation—the AI’s ability to connect the dots between your prompt, your history, and deep sonic attributes in its library. That’s the hard part. Getting it to generate *a* playlist is easy. Getting it to generate the *right* playlist, consistently? That’s the billion-dollar challenge.
