Two years ago, a Sónar+D panel called “Generating Panic?” asked whether AI was about to overwhelm music. By April 2026, Deezer had a number: around 75,000 fully AI-generated tracks a day, about 44% of all the new music arriving on the platform. The panel asked, the data answered. Then you
Alessandro Ravanetti
Two years ago, a Sónar+D panel called “Generating Panic?” asked whether AI was about to overwhelm music. By April 2026, Deezer had a number: around 75,000 fully AI-generated tracks a day, about 44% of all the new music arriving on the platform. The panel asked, the data answered.
Then you look at what people play. AI tracks are 1 to 3% of streams on Deezer, and the platform demonetises about 85% of them as suspected fraud. The uploads keep climbing. The listening has barely moved.
So at the Sónar+D 2026 pitch session in Barcelona this week, one thing stood out. Across the nine startup pitches, none was building an AI song generator.
This year this part of the wider Sónar festival moved into the Llotja de Mar, the neoclassical old maritime exchange in central Barcelona, and now runs as a non-profit under Fundación Sónar. The programme split into three strands, led by “AI & Music”, framed around a “post-AI” phase. By that it meant a time when using AI is already becoming “second nature”, with the attention shifting to the frictions that follow rather than the question of whether it belongs.
The same showed up in the startup pitches, hosted as ever by the Barcelona Music Tech Hub.
Two years ago the worry was the music itself: would a model write the next hit and put musicians out of work. The founders on stage this year had moved past that. They were pointing AI at the parts of the business that were never properly measured. The dancefloor. The royalty trail. The fan whose data you never owned. The value is moving off the song and into the plumbing.
Music Magnet: the dance floor as a data source
Music Magnet, based in Paris, calls itself an operating layer for music discovery. Its first product, DJ.pro, does something the industry mostly gave up on. It works out which tracks actually get played in clubs, where and when, using audio-recognition technology.
Last year on this stage I covered BEAM, which pulls together royalty and usage data that already exists. Music Magnet covers the previous step, to a signal nobody captured at all. The dancefloor is where genres are born, founder Frédéric Hérout argued, and almost none of it reaches the royalty system.
Hérout flagged a newer wrinkle, too: DJ crates are filling with AI-generated fake remixes, the same upload surge from streaming now reaching the dancefloor.
The recognition engine is built with Anthropic’s Claude, and the underlying technology, Auditio, secured Eurostars-3 funding with co-funding from Horizon Europe. A very European piece of plumbing.
2AM: the fan signal the organiser never held
2AM, also French, was co-founded by Marine Rivollet and Chloé Lavrat. Their pitch put a line on screen that captures a key part of what’s changing: “for decades, the people who run the show never held the data.”
Rivollet’s point: Meta and Ticketmaster have both rolled out AI agents in the past year, but, as she put it, each does it for itself: Ticketmaster for Ticketmaster, Meta for Meta. Who, she asked, is building one for the promoter?
2AM builds an Audience Experience Engine, a set of AI agents for festivals, venues and promoters. It gathers the fan signals that usually sit scattered across ticketing platforms and social channels, lands them in one place the organiser owns, then acts on them.
That covers a lone message at 2am and a crowd of 20,000 asking the same question in five languages on the night. The agent has already been tested at Rock en Seine, a 180,000-person festival near Paris.
It’s built in the EU and GDPR-native, which, for a product whose whole premise is fan data, actually matters to the buyers.
Yaarz: the money that leaks at the merch stand
The funding is following too. Yaarz, out of Station F in Paris, builds direct-to-fan commerce for live events: a platform and app for selling merchandise at the show, with audio sales feeding through to the official charts.
Founder Yohan Imakhoukhene raised €1 million from backers including Sony Music Entertainment France, the national music centre CNM and the public bank Bpifrance, and counts artists from Orelsan to Zaho de Sagazan among its users.
The merch table is one of the oldest revenue lines in music, and one of the leakiest. A major label putting money behind fixing it gives you hints about where the attention is going.
Parkinsonic: music as medicine, measured
Kate Berkita pitches Parkinsonic at Sónar+D 2026.
The pitch that stayed with me went furthest from music as a business. Parkinsonic, based in Germany, was built by Kate Berkita after her mother was diagnosed with Parkinson’s six years ago.
Someone with Parkinson’s might see a specialist only once or twice a year, while symptoms shift daily; one study found about 53% of patients are assessed twice a year or less. Parkinsonic aims at the days in between.
It’s an AI platform that assesses a patient’s neuromotor function, then generates personalised music sessions tuned to help them move. Rhythmic auditory cueing, using a steady beat to support the gait of people with Parkinson’s, is backed by years of clinical research; Parkinsonic wraps it in an assessment loop and a model that adapts to the individual. It sells to patients directly and to clinics.
Here the AI reads how someone moves and answers with rhythm to help them walk. That’s about as far from the 2024 panic as music tech gets.
The European through-line
Emanuele Sanfelici, co-founder and CEO of HAT Music, at Sónar+D 2026
One more, from the same session. Italy’s HAT Music, founded by Emanuele Sanfelici and based in Bologna, is an app that connects emerging artists with verified professionals: managers, producers, A&R, vocal coaches. It runs a Pitch Room where artists get feedback on tracks, and an in-app assistant, Sofia AI, to help with the matching.
Roughly 10,000 artists have used it to find a manager. It’s a fix for the fragmented middle of the industry, the artists sitting below the algorithm’s notice.
The big picture
Step back and the European thread is hard to miss. 2AM is GDPR-native by design, Music Magnet runs on Eurostars and Horizon money, Yaarz is backed by France’s public music funders. And several of the strongest founders this year were women, in a field that still needs many more.
In 2024 the worry was that AI would hollow music out from the core. Thankfully this didn’t happen and this year’s pitches point somewhere deeper: the most useful applications are showing up around the edges, in the layers the industry rarely measured. If there’s a trend to watch, it’s that one.
