The Question

A person wearing headphones with abstract musical waveforms and AI imagery

Spotify knows not just what songs you've listened to, but when — whether you play certain tracks when you're sad, whether your taste shifts on Sunday mornings, whether you skip the same bridge every time. Its algorithm learns from more than 700 million users, and its Discover Weekly playlist, launched in 2015, became one of its best-loved features.

But Spotify is still mainly a recommendation engine: it finds music that already exists. What is emerging is different — systems that create music for you, sometimes adapting as you listen. Endel, a Berlin-based startup founded in 2018, generates personalised ambient soundscapes that adapt to time of day, activity, and heart rate, and in 2019 it signed a distribution deal with Warner Music Group. But functional sound for focus, sleep, and exercise is a niche alongside mainstream music.

What the Evidence Shows

The AI music generation landscape has exploded in three years. Suno and Udio generate complete songs — vocals, instrumentation, lyrics — from a text prompt in seconds. Google and Meta have built text-to-music models of their own. AIVA, a Luxembourg-based company, has offered AI-composed scores since 2016. In a November 2025 Ipsos survey for Deezer of 9,000 people in eight countries, 97% could not tell fully AI-generated songs from human-made ones in a blind test.

The upload numbers are staggering — the listening numbers are not. Deezer reported in July 2026 that it was receiving about 90,000 fully AI-generated tracks a day, more than half of all new uploads on peak days. Yet such tracks accounted for only 1–3% of streams, and Deezer found that up to 85% of the streams they received in 2025 were fraudulent — bots inflating play counts to collect royalties. In the same survey, 80% of people said fully AI-generated music should be clearly labelled.

"Now that half of all daily uploads are AI-generated tracks, we are taking additional steps to safeguard the rights of artists and songwriters, while maintaining focus on music that fans actually love."

— Alexis Lanternier, CEO, Deezer, July 2026

The legal confrontation is turning into negotiation. In June 2024, Universal, Sony, and Warner sued Suno and Udio for copyright infringement over the music used to train their models. In October 2025, Universal settled with Udio and agreed to build a licensed AI music platform; Warner settled with both Udio and Suno, which agreed to launch new models trained on licensed music in 2026. As of April 2026, Sony was still in dispute, and talks between Universal and Suno had reportedly stalled.

AI can already flood streaming services with songs. Getting people to choose them over the music they love is a different matter.

Why This Is Happening

Streaming has already made music disposable. The shift from owning an album to streaming a playlist changed the listener's relationship with music: when you bought a record, you committed to it; when you stream, you skip after eight seconds. This has conditioned a generation to treat music as background infrastructure rather than a focal artistic experience — exactly what adaptive AI music is optimized to provide.

Wearable biometrics could create real-time feedback loops. The Apple Watch, Fitbit, and Garmin continuously track heart rate, movement, and sleep — precisely the data adaptive music systems need. A workout soundtrack that raises the tempo as your heart rate drops is technically possible today. Whether many people want it is another question.

The economics of AI-generated music threaten conventional production. A professionally produced song can cost thousands of dollars; an AI-generated one costs almost nothing. For stock-music composers and session musicians, the competition is immediate. But streaming payouts are tied to listening, and so far listeners mostly choose human artists.


What Could Happen

AI-personalized music becomes the dominant listening mode by 2035 Unlikely · 8%

Streaming services add personalised real-time soundscapes as features, and a generation raised on algorithmic curation adopts adaptive AI music as naturally as it adopted playlists over radio. The middle tier of working musicians shrinks, and AI becomes the default way most music is made and heard.

Legal rulings force a reset and slow adoption significantly Possible · 32%

Courts or licensing deals force AI music companies to retrain on licensed or public-domain material and share revenue with rights-holders. AI music grows, but on the labels' terms and more slowly.

A cultural backlash elevates human artistry as a premium Most likely · 60%

"Human-made" becomes a premium label, like organic food. AI dominates background and functional listening — sleep, focus, workouts, video soundtracks — while human artists dominate conscious, intentional listening. Labelling rules, which most listeners say they want, reinforce the divide. The industry bifurcates rather than transforms.

Our Assessment
We assign 8% probability — unlikely that AI-personalized music becomes the dominant form of music consumption by 2035. We have lowered this forecast from 71%. Our earlier version cited user, subscriber, and upload figures and a quote from Endel's co-founder that we could not verify. The real data show an explosion in AI uploads but very little genuine listening: fully AI-generated tracks make up only 1–3% of Deezer's streams, and most of those are fraudulent. Licensing deals will make AI tools more legitimate, but a split — AI for functional and background music, humans for the music people choose — is far likelier than the end of the fixed song.

What Can We Do

A musician performing live on stage, representing the irreplaceable human element in music

The transformation of music from fixed artifact to personalized, adaptive experience is underway. Whether you are a listener, a musician, or a rights-holder, there are choices to make about how it unfolds.

If you are a musician, build your irreplaceable self. The artists who thrive will be those whose fans are buying access to a human being, not just a sound. Live performance, direct fan relationships, distinctive visual identity, and genuine narrative authenticity are moats no AI can replicate. Musicians who treated their career as a product to be optimized will be displaced; those who treated it as a relationship to be deepened will find their audience more loyal than ever.

Read the terms of service on AI music tools. Most AI music generation platforms claim broad rights over content created on their platforms, and many train their models on user-generated content. Understanding what you are agreeing to when you use these tools — and what happens to the music you create with them — is essential for anyone creating professionally or semi-professionally.

Support copyright reform that protects creators. The legal frameworks governing AI training data are being determined right now. Organizations like the Featured Artists Coalition and the Songwriters Guild of America are advocating for creator rights in AI regulation. These debates will determine whether the value created by AI music flows predominantly to technology platforms or is meaningfully shared with the artists whose work trained the systems.

Be intentional about how you listen. Choosing to listen actively — to hear an album through, to attend a live performance, to engage with music as an aesthetic experience rather than ambient infrastructure — is both a personal and a cultural act. It signals to artists, platforms, and the industry what kind of music ecosystem you want to inhabit.