Google’s NotebookLM is the only AI tool I use almost every single day, and not just because it’s excellent at what it does (which it is). Its use cases go far beyond its core functions, which is exactly why I keep coming back. Beyond studying and work, I’ve used it for all sorts of weird (and, if I do say so myself, genius) experiments.

For instance, a couple of days ago, I realized I wanted to turn NotebookLM into a chaotic brainrot tutor to help me study. I’ve used NotebookLM to quickly speed through a TV show I can’t get into, bake the perfect tiramisu, and even pick up new habits. I also recently used NotebookLM to generate a Spotify Wrapped-like podcast (since the actual version wasn’t available in my region), and

that gave me a fresh idea: could NotebookLM help me make better playlists?

Why do I even want to use NotebookLM to create Spotify playlists?

It sounds strange, I know

Chances are, you’re wondering why on earth someone would use NotebookLM, a tool that’s best known as a research assistant out of all the tools there are, to create Spotify playlists. But hear me out: it’s the perfect tool for this usecase. A couple of days ago, I was wondering why there isn’t an AI feature that lets you generate Spotify playlists based on a prompt. I thought that’d be a fantastic AI feature, and one I’d genuinely use every day.

Upon digging around, I quickly discovered that Spotify is indeed testing a feature like this called AI Playlist. You describe a vibe, and it builds a playlist tailored to you. Exactly what I had in mind. Unfortunately, just like the Wrapped 2024 podcast, it’s not available in my region yet. Plus, from what I’ve heard, it’s still a little rough around the edges.

And even if it were available, I have very specific taste. Painfully specific. When I create playlists, over 90 percent of the tracks usually come from music I already listen to and gradually fall in love with over time. I like playlists that feel personal. The best way to ensure they’re tracks I enjoy is if they come directly from my own listening history.

Given NotebookLM is designed to work with data you feed it, I knew it’d work to create the AI Playlist Spotify. Once NotebookLM could access my listening history, in theory, I’d just need to describe the kind of playlist I wanted and let it surface the songs I actually listen to during those moments.

Setting it up is ridiculously easy

It takes seconds

As mentioned above, I wanted my playlists to consist of tracks I’ve already listened to before (but I enjoy most). Once NotebookLM could access my listening history, in theory, I’d just need to describe the kind of playlist I wanted and let it surface the songs I actually listen to during those moments. So, the very first thing I needed to do was give NotebookLM access to my listening data. I have Spotify connected to Last FM, so I decided to use its data.

If you’d like to do this as well and don’t have Last.FM, you can also download your listening history from Spotify directly. Just keep in mind it takes around five days to receive the files. I used a third-party website called BenjaminBenBen to export my listening data, which let me download all my data by simply entering my Last.fm username. A CSV file was generated with my listening data broken down into the following: Artist, Album, Track and Timestamp.

I then converted it into a PDF since NotebookLM doesn’t accept CSV files as sources, and added it as a source to a new NotebookLM notebook. That’s all the setup that was required! Now, all I needed to do from was this point onward was describe the kind of playlist I wanted to create, and NotebookLM would work its magic. For instance, I spend a little too much time getting ready in the morning, and I enjoy listening to some of my current favorite tracks as I take my sweet time.

My track rotation tends to stay pretty consistent for a few weeks before I suddenly latch onto a whole new set of songs. Most mornings, I’ll throw on one of my playlists and immediately start skipping until I land on whatever track I actually want to hear. Or I’ll just search it up directly because I cannot tolerate a wrong song first thing in the morning.

So, I wanted to create a playlist of the songs I’ve genuinely been listening to while getting ready recently, then mix in a couple of my other current favorites.

Since the listening history I exported from Last.fm included timestamps, all I had to do was tell NotebookLM what time I usually wake up and get ready. That way, it could filter my data and pull the tracks I tend to gravitate toward during those hours. Here’s the prompt I used:

Make me a playlist based on my morning getting-ready music. Filter my listening history to 7 to 8:30 ish AM, pick the songs I loop most often, and give me at least 30–50 tracks. Then throw in 10–15 other songs I’ve been obsessed with recently, even if I usually listen to them at other times.

Within seconds, the tool made a list of tracks I could add to my playlist.

I used an agentic browser to actually build the playlist

I let AI do the heavylifting

Once NotebookLM lists down the tracks I should add to a playlist, I have two ways to go about it. The most logical one is to, of course, manually add each song on Spotify like a normal person. But where’s the fun in that? I’ve really been into AI browsers lately, and I can’t imagine not using one in 2025.

Since agentic AI is capable of actually carrying out tasks for you, I prefer using it to automate redundant (and boring) tasks for me. Adding multiple songs to a playlist is a prime example of a redundant task I can outsource to AI. The playlist contained 81 tracks, so I figured this was the perfect moment to let AI do the heavy lifting.

Opera Neon is currently my primary browser, and its agentic AI capabilities are spectacular. Instead of spending 10 minutes searching and adding songs, I simply fed the playlist list to Opera’s agentic AI and asked it to create a new playlist on Spotify with all the tracks included. Because my Spotify account was already logged in, it handled the entire process for me.

Frankly, Neon did take a while to complete the task, and I could’ve done it a lot faster manually. But it kept working in the background while I got on with my other tasks for the day. Given this wasn’t time-sensitive, and I wasn’t exactly in the mood to drag-and-drop my way through 81 tracks, I didn’t mind the wait. By the time I checked back in, Neon was done, and my playlist was ready to go!

The possibilities are endless

I’ve used this method to generate playlists on days I don’t feel my best and need feel-good songs, when I’m pulling an all-nighter and want the music I usually reach for at 3 AM, and even on slow mornings where I want something soft and cozy. Every playlist feels personal, but I barely lift a finger.