Spotify's Plan For AI Generated Music, Podcasts, and Recommendations — With Gustav Söderström

Episode Summary

Executive Summary: Spotify’s Gustav Soderstrom framed AI as a powerful new creator tool that will expand music, podcast, and audiobook creation while raising urgent questions about copyright, compensation, and recommendation quality. He argued Spotify wants to support human creators, not generate content itself, and sees AI as making discovery more conversational, personalized, and scalable across formats.

Main Topics: AI music as a creator tool, not a replacement (Priority: 5/5): Soderstrom compared AI music to past creative tools like orchestras, synthesizers, and DAWs: each lowered technical barriers and expanded who could create. He argued AI should be treated as a bicycle for the mind that amplifies creativity, while acknowledging legal questions around training data and compensation. Spotify’s stance on synthetic content and copyright (Priority: 5/5): He said Spotify wants to allow legally created AI-assisted music on the platform, but not generate content itself. Spotify is building detection systems for derivative/copyrighted material and says the industry must develop a business model so creators are reimbursed in the AI era. NotebookLM and AI podcast generation (Priority: 4/5): Soderstrom described AI podcast hosts as impressive but best suited as a production tool rather than a full replacement. He argued NotebookLM succeeded because it turned source material into dialogue, which aligns with how humans consume podcasts, and it will expand the long tail of niche audio content. Recommendation systems becoming more interactive (Priority: 5/5): Spotify is moving from one-way algorithmic recommendations toward two-way conversations. Soderstrom said AI can make Spotify feel like a musical ‘friend,’ with AI DJ and AI playlisting giving users more context, more control, and eventually free-text feedback. User agency vs. algorithmic convenience (Priority: 4/5): He addressed criticism that Spotify can feel like it pushes users toward algorithmic listening. His answer: the company is re-emphasizing playlisting, adding controls, and trying to serve multiple user types—from album purists to casual listeners—without flattening taste. Podcasts, audiobooks, and unified audio distribution (Priority: 4/5): Spotify’s expansion into podcasts and audiobooks was justified by internal user demand, the need to improve discovery, and the advantage of keeping users in one app. Soderstrom argued software should adapt to content, not force users to switch apps for music, podcasts, or books. TikTok, social discovery, and culture distribution (Priority: 3/5): Soderstrom said TikTok and Instagram Reels are major top-of-funnel discovery channels for music, while Spotify is building integrations and editorial playlists to capture downstream listening. He also emphasized Spotify’s role in social sharing and joint listening.

Key Arguments: AI music is a continuum, not a binary; many major artists already use AI for parts of songs, so the industry should focus on how tools are used rather than whether they exist. Spotify’s mission is to support creators; if AI-generated music is legal and creators are compensated, Spotify should let people listen to it. The real problem with AI content is not just copyright, but the absence of a business model that pays the humans whose work and participation make the ecosystem possible. NotebookLM-style AI content works because it transforms source material into dialogue, which is more engaging than monologue and fits the podcast medium better. AI will dramatically expand the amount of niche content available, which makes recommendation systems even more important, not less. Recommendation systems should become more conversational so users can explain what they want instead of only signaling through skips and clicks. Spotify should balance convenience with agency; going to zero user investment may improve short-term engagement but reduces trust and participatory feeling. Human-led identity and fandom still matter, so AI may dominate utility use cases like study music or ambient audio, while artists and human connection retain special value. Spotify’s multi-format strategy makes sense because users want one app that adapts behavior—skip buttons, chapters, discovery—based on whether they are hearing music, podcasts, or audiobooks. Social platforms like TikTok strongly shape music culture, but Spotify aims to integrate with those platforms while also being a place where culture can happen directly.

Data Points: Spotify catalog growth: from tens of millions of tracks to hundreds of millions of tracks - Soderstrom described the music library expansion since Spotify’s early days AI DJ usage: bigger than Discover Weekly usage for people who use it - He said AI DJ has been successful and its usage exceeds Discover Weekly among its users Podcast listening format: an hour or more is a common podcast session - Used to explain why podcast discovery requires stronger previews than music Music listening with others: double digit percentage - He said surveys show a substantial share of listening happens socially, not purely alone Recommendation hit rate example: 1 in 20 - He used this as an example of low hit rate when exploring very broad new music Podcast end-of-episode dropoff: big drop at about 90 something percent - He noted most listeners fall off near the end and that Spotify accounts for end credits NotebookLM format: dialogue - He said Google’s breakthrough was turning source material into a two-host conversation AI playlisting feedback loop: yes/no/free-text prompting - He described a new Spotify experience where users can refine playlist requests in plain language

Pivotal Quotes: "I think of them as tools." — Gustav Soderstrom: Explaining his overall philosophy on AI music and creative software "Our goal is to cater to everyone." — Gustav Soderstrom: Discussing Spotify’s need to serve album listeners, playlist users, radio-style users, and others "I think in 2024, the user should not adapt the software to the content. I think in 2024, the software should adapt to the content." — Gustav Soderstrom: Justifying Spotify’s unified app strategy across music, podcasts, and audiobooks

Implications: Spotify is positioning AI as a creator-enabling layer while defending human creativity, compensation, and discovery. Expect more conversational recommendations, more niche audio, and more pressure on platforms to balance automation with user control.