Artificial intelligence is no longer just assisting developers at Spotify — in some cases, it’s effectively taken over the keyboard.
During the company’s fourth-quarter earnings discussion, co-CEO Gustav Söderström revealed a striking detail: some of Spotify’s strongest engineers haven’t manually written code since December. Instead, they’ve been relying on AI systems to handle much of the heavy lifting in software development.
This shift comes amid an especially productive year for the streaming giant. Over the course of 2025, Spotify introduced more than 50 updates and new features to its app. Among the most recent additions are AI-powered Prompted Playlists, Page Match for audiobooks, and a feature called About This Song — all designed to deepen user engagement and personalize the listening experience.
Behind the scenes, much of this accelerated output is powered by an internal AI-driven tool known as “Honk.” The system enables engineers to deploy and refine code remotely using generative AI technology, including tools like Claude. The process, according to Söderström, can be remarkably seamless.
He described a scenario in which an engineer commuting to work could message an AI assistant through Slack, requesting a bug fix or a new feature for the iOS app. By the time that commute ends, the AI may have already implemented the change and sent back an updated build for review and approval — all before the engineer steps into the office.
Spotify credits this workflow with dramatically shortening development cycles and speeding up product releases. Leadership views the current stage of AI integration not as a peak, but as an early chapter in a much larger transformation.
Beyond productivity gains, the company believes it holds a competitive edge in data. Unlike general-purpose large language models trained on widely available sources, Spotify’s strength lies in its uniquely nuanced music data. Musical taste, Söderström pointed out, rarely has one “correct” answer. Workout playlists, for example, vary widely by region and preference — from hip-hop in the U.S. to EDM across parts of Europe, and heavy metal in Scandinavia.
Spotify says it is continuously refining this large-scale, preference-based dataset to improve recommendations and personalization.
As for AI-generated music, the platform allows artists and labels to disclose how tracks were created through metadata tags. At the same time, Spotify says it remains vigilant in filtering out spam and low-quality content, aiming to balance innovation with platform integrity.
