Why Spotify AI more than music is secret to keeping listeners

Streaming music apps have been nudging customers into the bogus intelligence period with a restricted observe file of success. But AI-based recommendation tools from Apple, Amazon and pure-play streaming firm Spotify are shifting forward, with Spotify’s newest strategy to the way forward for private music discovery leaning into the AI immediate in a number of codecs. Consultants say these tech investments could also be essential to Spotify’s skill to construct a moat round its enterprise because the core enter, music, turns into commoditized throughout the streaming apps.

A brand new ChatGPT integration rolled out recently by Spotify permits customers to attach their accounts on to OpenAI’s generative AI chatbot. The launch is nice for OpenAI in its broader effort to show ChatGPT right into a platform for third-party apps that perform inside conversations. For Spotify, it is a wager that personalised music and podcast suggestions will probably be improved via the now-familiar format of chatting with an AI and letting it know what you need.  

Spotify customers can ask for songs, artists, albums, playlists or podcast episodes by temper, style or matter. Outcomes floor inside ChatGPT and open within the Spotify app for playback. Customers can work together with the advice, and provide specificity past what is feasible with a basic “like/dislike” suggestions choice. 

In response to a Spotify spokesperson, the prompts are “a possibility to uncover new tracks or revisit previous favorites, or lengthen a ChatGPT dialog with a soundtrack that matches the second.” 

Spotify mentioned the combination is opt-in and that customers can disconnect at any time. It additionally mentioned it won’t share music or podcast content material with OpenAI for coaching functions — addressing trade issues round AI and copyrighted materials.

Spotify additionally just lately rolled out its Prompted Playlist characteristic contained in the streaming music app, a “vibey” feature that enables customers to faucet into a sense or reminiscence to be able to construct a customized combine.

The rival streaming providers related to the massive tech gamers are exploring comparable AI options.

Apple has been progressively layering AI into Apple Music. The “Playlist Playground” beta characteristic is closest to what Spotify is doing, because it additionally focuses on chat-based AI interplay that enables customers to tweak suggestions by way of chat. Apple recently introduced AutoMix, which makes use of AI to investigate songs and mechanically mix tracks by matching tempo and beats, eliminating silence between songs, including crossfades and so forth. The corporate has additionally rolled out machine-learning instruments resembling lyric translation and pronunciation features. 

Amazon Music has supplied a prompt-based playlist characteristic referred to as Maestro since mid-2024, which permits listeners to generate playlists utilizing textual content descriptions and even emojis. It stays in beta testing somewhat than full launch.

Spotify executives have repeatedly described AI as central to the platform’s subscriber stickiness technique. On a recent earnings call, management advised traders that enhancements in AI-driven discovery are central to maintaining customers engaged with the platform. “Our investments into personalization and AI are paying off,” mentioned Alex Norström, co-chief government officer. “It means individuals are spending extra days in a month with us and throughout extra moments,” he mentioned.

Spotify’s interactive iDJ characteristic, launched in 2023, has roughly 90 million subscribers utilizing it as of the latest earnings report, with customers racking up over 4 billion hours of time spent on the app. Norström mentioned Prompted Playlists as “immediately taken off with energy customers.”

“If iDJ is the chat interface to Spotify, the place you possibly can speak casually, Prompted Playlists is the Deep Analysis mode of Spotify,” he mentioned. “It helps you to describe and set guidelines in your personal personalised playlists – actually writing your individual algorithm. … There’s nothing else prefer it.”

Music catalogs commoditized, AI generates hundreds of thousands of songs

In response to analysts who cowl Spotify, the chief hype about AI might have to be the fact sooner somewhat than later for the corporate. Although there are instances of musicians pulling music from particular apps every now and then as destructive headline points seem — it happened to Spotify most just lately on account of its founder and former CEO Daniel Ek’s defense tech investments — rivals together with Apple Music, Amazon Music and YouTube Music provide largely overlapping catalogs and more and more refined advice engines.

“The catalogs at Amazon, Apple and YouTube are comparable — almost an identical songs — to Spotify, similar to Bing and Edge are almost an identical to Google,” mentioned Michael Pachter, senior advisor, digital media, sports activities & leisure at Wedbush Securities, who lined the streaming trade as analysis analyst for a few years. (Wedbush has by no means had a score on Spotify shares particularly.)

Whereas Google’s search enterprise faces its personal AI risk, Pachter mentioned it’s also one of the best mannequin for Spotify to look to when it comes to learn how to preserve a person edge. “Google managed to widen its moat by providing plenty of options that make the service stickier, together with remembering my bank card and password information. I can not even conceive of switching from Google Search, and I believe that’s what Spotify is attempting to ascertain,” he mentioned.

Switching prices could also be refined, however they are often important. Customers construct libraries, curate playlists and prepare algorithms over years. Every further integration, whether or not with a automobile dashboard, a voice assistant or now an AI chatbot — Spotify says it now connects to over 2,000 gadget varieties — can additional entrench the ecosystem.

“I count on this ChatGPT integration will probably be broadly utilized by Spotify customers and wildly profitable,” Pachter mentioned. “Others might attempt to do the identical factor, however the switching prices develop each time you take the time to construct your playlists on Spotify, and that is what they’re relying on,” he mentioned.

Apple Music and different third-party apps do provide instruments to export playlists when subscribers search to alter music providers.

Others on Wall Road are much less satisfied than Pachter, however did come away from the latest quarter extra optimistic on the Spotify AI story and fewer fearful in regards to the dangers it faces from AI music creation instruments disrupting platforms like its personal. Spotify’s inventory value has slumped up to now 12 months by shut to twenty%, although the inventory has carried out very strongly since its 2018 IPO.

“Spotify addressed this concern head‑on, arguing that AI helps somewhat than undermines its strategic place. By leaning into personalization, product innovation, and scale benefits, Spotify seems positioned to make use of AI to strengthen its platform, although the tempo of adoption and trade alignment will stay key variables,” wrote Financial institution of America’s analysis workforce, which charges the shares a purchase, in a February word after the latest earnings.

Gustav Söderström, Spotify co-CEO, mentioned throughout that earnings name that constructing a music app that customers can speak to, and that absolutely understands every listener, will shift listening “from a passive expertise to an interactive one.”

Mark Mulligan, managing director and analyst at MIDiA Analysis, a analysis agency that tracks the music market, says AI goes to be integral to streaming music conduct, however he’s much less satisfied that the excellence between interactive and passive being made by Spotify is the probably consequence.

“Streaming music consumption has bifurcated between passive and energetic,” Mulligan mentioned. “However this doesn’t imply that the viewers has cut up in two — everybody has. Even probably the most energetic music listeners spend greater than half their time listening passively.”

Actually, he says the broader development is towards extra passive consumption via curated playlists, and likewise by way of options resembling artist radio stations and AI DJs. “The course of journey is in the direction of extra passive listening,” Mulligan mentioned. Agentic options might signify a compromise, “a center floor between passive and energetic listening,” he mentioned. “It permits the person to expend a small quantity of ‘lean ahead’ effort in return for a considerable amount of ‘lean again’ listening.”

Typing an in depth immediate into ChatGPT might really feel energetic, however Mulligan says “the extra that the algorithm learns in regards to the listener’s conduct and tastes, the higher its suggestions get and due to this fact the much less the person has to lean in, thus shifting the needle even additional in the direction of passive listening.”

On this AI interface-first mannequin of streaming, the underlying content material is necessary, however issues much less to what in the end makes the person really feel rewarded. For instance, a capability to explicitly exclude artists or slender by subgenre might make AI-assisted discovery really feel extra tailor-made than conventional algorithmic playlists. If a listener enjoys some Nineteen Eighties rock bands, resembling Bon Jovi and Weapons N’ Roses, however dislikes others from the identical period, it is simpler to filter. Spotify can usually predict that if you happen to like A and most of the people who like A additionally like B, then you definately’ll most likely like B, however that does not essentially signify how person tastes are expressed. “With GPT, I might say ‘no Def Leppard’ and my lists could be scrubbed of them,” Pachter mentioned.

Any predictions in regards to the influence of AI on music, as with all trade, stay educated guesswork. However it’s already very clear that AI is having an influence on the concept of a music catalog itself. In response to a latest Rothschild & Co Redburn report, text-to-music platforms like Suno are reportedly producing round seven million songs per day, roughly equal to Spotify’s whole pre-AI catalogue each fortnight. “This can be a deluge,” its analyst Ed Vyvyan acknowledged.

Söderström hinted that it is the dataset but to come back and never the deep tracks already within the stack that matter most to the long run. “We’re constructing a dataset that by no means existed,” he mentioned on the latest earnings name. “We’ve had the song-to-song dataset, however nobody had the language-to-song dataset. … It’s possible you’ll suppose it’s a canonical dataset, which means there’s a factual reply to, for instance, what’s exercise music? There isn’t a factual reply to what’s exercise music. … on common, for an American, it’s often hip hop. For a European, it’s often EDM. For a lot of Scandinavians, it’s one thing like heavy metallic and even dying metallic. Then once more, for lots of People, hundreds of thousands at the least, it’s also dying metallic.”

“You can’t simply have an LLM commoditize it as a reality, the best way you possibly can commoditize Wikipedia,” he mentioned. “You really have to have many, many lots of of hundreds of thousands of listeners the world over’s markets continually telling you what it means for that particular particular person.”