"People may not always be able to hear whether a person is attached to the music, but they still care whether one is."
Guest post by Bas Grasmayer, who leads growth strategy at Vandall
In July, Deezer announced that more than 50% of new uploads were fully AI in June 2026. Across the month, around 90,000 AI-generated tracks were being delivered to the platform each day. Up from 10,000 only 18 months prior.
Synthetic music can be produced and distributed at a speed that human creators will never match, but upload volume is not the same as cultural impact.
Fully AI-generated tracks still account for only a small share of listening on streaming (between 1-3% of plays on Deezer), and much of that activity has been linked to streaming fraud. These songs are fast to create and fast to move on from and are slung at the algorithm without an artist building a world around them or a team working a release plan.
The current tipping point may be easy to dismiss because so little of the AI music attracts meaningful attention. But the fact that it’s not happening yet, doesn’t mean a playbook won’t emerge to establish synthetic music as a serious competitor for our attention.
People care who made the music
As music streaming rose in popularity in the 2010s, so did live music and physical formats like vinyl and cassettes. The IFPI reported global physical revenues up 8% in 2025, with vinyl up 13.7%. Luminate found live music spending still climbing despite ticket prices that keep rising. People are willing to pay a premium for authenticity.
Another Deezer-commissioned survey found that 97% of listeners failed to distinguish fully AI-generated music from human-made music in a blind test. At the same time, 80% said fully AI-generated music should be clearly labelled.
Songs often carry relationships, biographies, scenes, memories and stories. People may not always be able to hear whether a person is attached to the music, but they still care whether one is.
The problem isn’t creation
The response to AI flooding the streaming services is not complicated. Keep putting people in rooms together. Keep running the writing camps. Keep giving artists however long they need, though sometimes it can be helpful to tell someone when the track's done, also.
The problem is what happens afterwards. A song can be written in a week and spend the following six months trapped in admin. Files are spread across inboxes, drives, and expiring links. Decisions sit in unconnected WhatsApp threads, emails, and voice notes.
Music industry outsiders are often surprised by how normal this is. Paperwork, often seen as the boring part, gets relegated to ‘later’ only to become urgent when the song starts earning, rather than being treated as part of the collaboration itself.
This becomes a liability in the age of AI when agents can create, release, and test new synthetic music continuously.
Speeding up coordination
The next generation of successful music companies may not look like the large, vertically organized labels that dominated the previous century.
They may look more like small, highly capable core teams coordinating much larger networks of artists, writers, producers, managers and specialist partners.
We have already seen early versions of this model in dance music. UKF began as two YouTube channels in 2009 before expanding into releases, editorial and events. Monstercat launched on YouTube in 2011, building a digital-first label around a large network of producers and a high release cadence. Spinnin’ Records was also early on YouTube, launching its channel in 2007 and building it into a major distribution and discovery engine: by 2019 it was releasing 15 to 18 tracks a week and had more than 26 million subscribers. Warner acquired the company in 2017 for a reported $100m+. These teams showed that smaller, networked music companies could move faster by embracing new digital tools and building around how music was now being made, shared and discovered.
The next version will go further.
Five years from now, the strongest music companies may still make songs in much the same way they do today. People will meet, exchange ideas, disagree, improvise and take their time. What changes is everything surrounding that process.
Rights information will be captured while the work happens. Files, decisions, agreements and plans will live in one place. Systems will surface what is missing before it becomes a problem. Release checklists will improve with every campaign. Teams will be able to see what worked, learn from it and apply those lessons to the next project. Commercial planning will become part of the workflow rather than a separate scramble that begins once the music is finished.
Ironically, AI may be one of the things that makes this possible. Not necessarily by making the music, but by taking more of the coordination, planning, organization and repetitive work off people's plates.
Some organizations already work that way, but it will become less optional in the next few years. Orgs that thrive will look like a small group of people with the ability to coordinate a much larger network than in the past without adding the same amount of administrative overhead.
The bad outcome is an industry in which human music remains trapped in fragmented systems while synthetic music becomes faster, cheaper and better organized. The cost shows up in teams that stay bloated, workflows that do not scale, and people spending more time coordinating the work than improving it.
The better outcome is that teams become exceptionally good at supporting the things machines cannot reproduce: trust, relationships, judgement, and shared history, while using technology to remove as much friction as possible from everything around them.
We don't need to outrun AI; we just need to stop tripping over ourselves.
Bas Grasmayer has led product at music streaming services and platform strategy at COLORS. He founded the MUSIC x newsletter and works with Vandall on collaboration infrastructure for the music industry.
