Spotify's AI DJ Fails Classical Music, Exposing Broader Industry Gaps

The Illiterate AI: A Stumbling Block for Centuries of Music

When author and programmer Charles Petzold asked Spotify's AI DJ to "play Beethoven's 7th Symphony," he expected a coherent performance of the four-movement masterpiece. Instead, the feature, launched in beta and promoted as a smart music discovery tool, delivered a baffling sequence. It played only the famous second movement, then abruptly switched to unrelated works by Mascagni and Shostakovich.

This failure wasn't a one-off glitch. Petzold's documented experiments in February 2026 revealed a pattern of profound misunderstanding. Even explicit prompts like "play all four movements of Beethoven's 7th Symphony in numerical order" yielded incorrect symphonies and jumbled movements from different recordings. The AI seemed incapable of grasping a foundational concept of Western classical music: that a composition is often a sequenced collection of movements.

The root cause lies in decades-old digital music infrastructure. As Petzold notes, metadata for digital tracks is built on a pop-song model of Artist, Album, and Song. For instrumental works, calling a movement a "song" is a misnomer. More critically, there is no standard metadata field linking individual movement tracks back to their parent composition, leaving AI systems blind to the intended structure.

A Flood of AI-Generated Music Meets Broken Discovery

Spotify's struggle with classical music curation occurs against a backdrop of an AI music explosion. According to a Billboard report, AI music startup Suno generates a staggering 7 million songs per day—equivalent to Spotify's entire catalog every two weeks. Deezer research suggests 97% of listeners cannot differentiate AI from human-made music.

AI-generated acts have already made chart inroads. Suno-created artist Monet topped Billboard's Hot Gospel Songs chart, while another AI persona, Ray, reached number one on the Gospel Digital Song Sales chart. The sheer volume and improving quality of this content make functional discovery tools more critical than ever.

If an AI cannot reliably retrieve a canonical work like a Beethoven symphony, how will it navigate an ocean of AI-generated tracks? This failure exposes a risk: discovery algorithms may prioritize easily categorized, pop-structure AI music over complex, metadata-poor legacy works, potentially burying centuries of musical tradition.

Industry Scramble: Detection, Disclosure, and User Control

The industry is reacting to the AI wave with a mix of defensive and user-centric features. Spotify has removed tracks that impersonate human artists without consent and begun adding AI disclosure tags in song credits. Rival Apple Music now mandates labels disclose AI use in artwork, songs, composition, and videos.

French service Deezer has rolled out an AI detection tool to flag generative AI tracks. Simultaneously, platforms are giving users more control. In March 2026, Spotify announced a "Taste Profile" editor, first for Premium users in New Zealand. This feature, unveiled by co-CEO Gustav Söderström at SXSW, allows users to see and directly shape the data model powering their recommendations.

This move acknowledges a long-standing user complaint: opaque algorithms that don't reflect true interests, especially in shared account scenarios. It represents a shift from a purely black-box AI model to a more collaborative, editable system of user preference.

The Core Conflict: Profit vs. Cultural Stewardship

Petzold's experience points to a sobering conclusion. "There is nothing less consequential to corporate profits than the preservation of the western musical tradition," he writes. The economic incentive to perfect AI for the vast pop and emerging AI-generated music markets likely far outweighs the impetus to solve niche classical metadata puzzles.

The lawsuits against AI music generators Udio and Suno by major labels and artists, noted in a Forbes analysis, focus on copyright infringement. They seek to set the "rules of the road" for commercial use. However, they do not address the fundamental structural ignorance of music that Petzold's experiment highlights.

For marketers, as discussed in Forbes, the copyright resolution will open doors to using AI music. However, audience sentiment and the ability of platforms to contextually recommend such music appropriately remain open questions, complicated by these underlying technical deficiencies.

A Question of Intelligence and Responsibility

The episode forces a reevaluation of what "intelligence" means in an AI context. An system that can generate convincing music but cannot retrieve a basic symphony as intended is displaying a narrow, brittle form of competence. The failure is ultimately one of programming and data architecture, not of a mysterious AI entity.

As Petzold asks, if an AI cannot grasp basic musical concepts, how credible are claims that it can compose? The problem is fixable—by enriching metadata schemas and training AI on compositional structures—but requires deliberate effort and investment.

The coming years will test whether music streaming platforms, now central cultural archives, view the integrity of their catalogs as a core responsibility. As AI reshapes music creation, the ability to understand, organize, and respectfully serve all musical traditions remains a critical benchmark for the technology's true maturity.