Spotify Wrapped’s AI Podcast and the Joyless Direction of Streaming

This year’s Spotify Wrapped brought a strange surprise. The app’s annual summary included more than just the usual candy-colored data about which artists, songs, and podcasts users streamed the most. It included a “Wrapped AI Podcast” feature that leveraged Google’s NotebookLM technology to verbally reflect users’ listening habits.

This bizarre addition is more than just an audio companion to the information presented in Wrapped. Two NPR-influenced artificial intelligence voices go through your listening highlights in a conversational (if still robotic) manner, touching on your “moods,” “vibes,” and interests throughout the year. It feels like listening to a doctor go over the results of your blood test and a psychic guess vague facts about your life.

For example, based on the movie soundtracks I was consuming, my “hosts” suggested that I adopt a “bolder” mood in September. “The challenger Soundtrack and Ludwig Göransson’s work Oppenheimer I wonder if you have seen some intense films during this time,” one of my “moderators” speculated. That wasn’t entirely untrue. But I also wasn’t sure why I needed to be told this information about myself.

The purpose of this new feature is not entirely clear. The podcast element feels even less inspired than some of the other strange personalization features the music streamer has introduced in the past, from an AI DJ that greets users by name to “sound cities” where the listener supposedly lives, right down to this year’s incoherent labels that represent different moods (such as “Pink Pilates Princess Vogue Pop”). These features are presented as attempts to connect with users on an intimate level, but for all their absurdity, they have largely proven to be great marketing strategies.

“The goal of Wrapped in particular is brand virality,” says Glenn McDonald, former data alchemist at Spotify. “There’s not much in the way of data storytelling this year.”

Overall, the Spotify listening experience has never felt stranger.

Spotify’s personalization features are getting weirder

Launched in 2008, the app has long been appreciated—and lamented—for its custom, algorithmically generated playlists and countless compilations designed for every mood, task, and time of day. However, newer personalization features with a focus on AI technology have gotten a bit out of control. While its technology is designed to understand users’ tastes, it is this very understanding that forever keeps users in a repetitive comfort zone of songs, artists and genres.

In 2023, the app launched an AI-powered DJ service called “DJ X”, currently available as “DJ: Wrapped”. Using a selection of users’ most played songs and forgotten favorites, the feature mimics the FM radio experience – just without the crucial element of discovery and a human guide. The AI ​​DJ – who uses the recreated voice of Spotify’s Head of Cultural Partnerships, Xavier “X” Jernigan – offers a vague introduction between songs: either basic facts about the artists or what mood they think the next song will evoke. It feels more like an attempt at a guided relaxation ritual than an adventurous listening experience. (Of course, users also have the option to skip songs.)

Another AI-powered tool Spotify introduced earlier this year is called AI Playlist. Users can either enter a description of the type of playlist they want to listen to or select a prompt and the generative AI will create it instantly. While you can request any mix you want, the prompts seem to encourage the user to listen to music they are already familiar with. For example, the site currently suggests that users try asking the tool to “rank their most popular Wrapped songs by release date” or “give them a playlist of the genre they’ve listened to most.”

Some of these functions have become much more complicated. This has made them effective marketing tools. Back in February, the app launched the trippy, cosmic-themed tool “Song Psychic,” which mimics a Magic 8 Ball. Users select ready-made questions about their future and receive a (well-known) song in return. The “Daylist” function, which quickly spread on social media at the beginning of the year, is better known. The “hyper-personalized” playlist changes throughout the day to supposedly reflect the user’s mood based on their listening habits at certain times of the day. It primarily attracts attention for its AI-generated playlist titles that look like word salad (e.g. “Rage People Pleaser Wednesday Night”).

It’s unclear whether the “AI Podcast” will be a mainstay of Spotify Wrapped or whether Google NotebookLM will be further integrated into the app. The “Research Assistant” allows users to receive summaries and additional information about documents using Google’s Gemini chatbot. However, when it comes to the podcast, its function seems a bit unnecessary for the user. It’s mostly a testament to the power of Google NotebookLM rather than something that significantly improves the user experience.

Are all these AI tools just making us self-centered?

These features not only show how well Spotify knows you, but also how predictable your behavior is in the app. These inventions can be fun to play around with. Sometimes they are exactly what users need at the moment. But to what end do streaming services need to continue to amplify and accommodate our own moods and preferences?

McDonald, who now runs the music directory Every Noise At Once, says it’s a normal tendency for users to listen to music they already like. “There are certainly retention metrics that say that people, not surprisingly, like to hear the things that they like,” he says. “But unless you do really adventurous experiments to find out what people might respond to, you really don’t know.”

During his decade at Spotify, McDonald said the streaming service never focused on “what it takes to make people curious.”

While Spotify has a popular “Discover Weekly” playlist that features new songs and artists based on users’ tastes, there are less easily recognizable features in the app’s interface designed for exploration. Compare that to playlists the app prominently features, like “Your Time Capsule,” “On Repeat,” and “Repeat Rewind,” which bring all users back into their listening comfort zone.

In all these efforts to get to know listeners, music is ignored as an interactive art form. These Spotify features suggest that music is all about your own enjoyment and ease, rather than something to engage with, seek out, and even criticize. The user-focused promotion of AI reinforces these habits. The audience should be happy that a robot knows everything about them, but without asking them to expand or change. All in all, these AI-powered features promote a narcissistic and, frankly, boring approach to art.

“The argument for streaming services is that all the world’s music is now available to you,” says McDonald. “But if you only listen to what you already know, it’s like giving someone a teleporter and they just use it to teleport themselves home.”