I thought I'd hate Spotify's AI playlists, but I don't

I like to think that I have pretty good taste in music. I’d go so far as to say I’m good at getting my friends to listen to new music, whether it’s through a playlist at a party or by attaching a new song to an Instagram post. However, with such a well-trained Spotify algorithm, even I find myself in a music rut on occasion. I’ll bounce from my pop mix to my Jason Isbell mix to my upbeat mix and somehow find quite a few of the same songs on each one.

So, when Spotify introduced Prompted Playlists, I figured it was a good chance to try something new. After all, if I could lead the algorithm to the vibe I was trying to capture, it would have to give me new music. At worst, I could simply direct it to introduce me to new tunes as part of my prompt. So, I opened my create tab, started a new playlist, and here’s what I got.

It pays to be precise

Before I get to the results of my first few AI-generated playlists, I should probably clarify the concept of Prompted Playlists a little better. After all, there’s at least some AI involved in everything you already get from Spotify, so how is this any different? Well, for starters, Prompted Playlists work a little bit more like text-based image generators. You type in exactly what you’re after, and Spotify does its best to meet those requirements.

Unlike an image that might come with odd artifacts like an extra finger or creepy teeth, the Prompted Playlists pull from the platform’s expansive library. In other words, at least you know there are some guardrails in place to keep you from getting anything too unusual. Then, once you have your playlist, you can rename it, rework your prompt, or ask Spotify to update the music that’s included — not far off from the idea of editing or regenerating an image prompt.

And now that you have the basics, let’s talk about the easiest, most obvious mistake I made with my very first playlist. I’ll admit I didn’t know just how specific I’d have to be, so I typed, “make me a slow, wintery playlist that’s full of indie music and perfect for shoveling snow.” Pretty simple, pretty straightforward, and I figured it would fill out with music I already knew and liked.

Well, it didn’t.

Instead, I got a mix of Italian and Vietnamese tunes that technically fit what I asked for but might have strayed too far off the beaten path. I listened for a few minutes, but couldn’t relate to lyrics I couldn’t understand, even if the instrumentals were just what I like. So, I jumped right in and began to rework my prompt. And, by that, I mean that I just specified that I wanted familiar music to be included.

One generation later, and I had about an hour of tunes that fit the bill and allowed me to sing along while I dug my car out from more than eight inches of snow — plus whatever the plow trucks had pushed my way. Better yet, Spotify gave a reason for each of the 20 songs on the playlist, highlighting everything from certain slow tempos to instruments that evoke images of winter. And Phoebe Bridgers. Like, a lot of Phoebe Bridgers.

Stop worrying, start experimenting

Luckily for Spotify, I like Phoebe Bridgers. In fact, I liked pretty much everything the algorithm put into my 20-song set, which isn’t exactly what I expected. Although I wanted some familiar jams, I didn’t want everything to feel like I’d heard it before. So this time, instead of reworking my prompt, I decided to start over. After all, there was only so much time I could spend shoveling before my car was all the way clean.

And, with the world back at my fingertips, I decided to start a playlist for what I spend most of my free time doing: running. I know I won’t be running quickly for a while, given the snowy conditions, but I asked for a mix that evokes warm summer days on trails with plenty of adventure, while also clarifying that I wanted familiar music as a base, with new songs added throughout. This time, when I pressed to generate the playlist, I got much closer to nailing it on my first try.

Once again, I got somewhere around 20 songs, each of which with an explanation, and I’d heard about 15 of them before. Surprisingly, most of the new songs were from the same band, Kingfishr, but after a few miles of listening, I’d say they fit right in. If there’s one issue with this mix, though, it’s that Spotify’s beats-per-minute range from a pretty casual 115 to a very, very fast 190, making it challenging to keep one pace all the way through.

Then, out of sheer curiosity, I asked Spotify for a jazzy, two-hour playlist for happy hour. Honestly, I don’t listen to much jazz outside of Laufey, so I wasn’t sure how this would go. Thankfully, the algorithm returned two Laufey songs out of the 25 it chose for me, which felt like a decent base of familiarity. It was enough of a head start for me to give the rest of the playlist a try, and I think I finally found the perfect balance of new music and old favorites.

Unlike the first playlist, which was totally unfamiliar, I’d managed to create a mix that aligned with the taste I wanted while also expanding my horizons with new sounds I could share with my friends. And, now that I think I’ve figured out at least a little bit of the algorithm, I’m excited to keep experimenting with different genres and requirements.