Qobuz has just done something about artificial intelligence that I wish more of the music and broadcasting industries would understand: it stopped trying to turn AI into a religious argument and started giving the audience information. The independent streaming company is now identifying music its technology determines to be AI-generated, placing a visible designation on those releases and telling listeners, essentially, here is what we believe this is and now you can decide whether you want to hear it. At the same time, Qobuz is keeping its editorial playlists, featured releases and music discovery operation firmly in human hands.
THAT is an AI policy.
Not panic.
Not pretending the technology does not exist.
Not screaming that every computer is coming to murder creativity.
And not quietly stuffing synthetic records into playlists and hoping nobody asks where the hell the artist came from.
TRANSPARENCY.
What a concept.
QOBUZ IS NOT BANNING AI
That distinction is important.
Music identified by Qobuz as AI-generated can still exist on the service.
The platform is not positioning itself as some technological monastery where algorithms are forbidden to enter.
Instead, Qobuz says its detection system identifies fully AI-generated content, places a visible badge on qualifying releases and removes that material from the places where Qobuz itself actively recommends music to listeners.
In other words:
YOU CAN EXIST HERE
BUT WE ARE NOT GOING TO PRETEND WE DON’T KNOW WHAT YOU ARE
I love that.
Because this is the conversation the music industry desperately needs.
LET THE LISTENER KNOW
That’s it.
If I am listening to a human singer who spent three years writing a record, tell me who made it.
If I am listening to something created almost entirely by a generative system, tell me that too.
Then let me decide.
Maybe I love it.
Maybe I hate it.
Maybe I could not possibly care less how it was made.
Fine.
But don’t remove the information necessary for me to make that decision.
Qobuz Deputy CEO Georges Fornay framed the company’s position around that basic idea when announcing the September rollout: detection alone is not enough; listeners also need to be able to recognize synthetic material when they encounter it.
That is not anti-AI.
That is pro-consumer.
AND THE NUMBERS ARE WILD
Here is where this story goes from philosophy to:
OH DAMN
Qobuz says only 0.38 percent of its current streams come from tracks the company has identified as AI-generated.
That is tiny.
But look underneath it.
Qobuz says 60 percent of streams involving tracks it has identified as AI-generated are currently being classified as fraudulent by its anti-fraud systems and excluded from royalty reporting.
Sixty percent.
Now we have a much bigger issue than whether somebody likes an AI song.
We are talking about the integrity of the music economy.
Who is listening?
Were they real?
Who gets paid?
Who gets crowded out?
What happens when industrially generated music and artificial streaming activity begin feeding each other?
Suddenly the argument isn’t:
DO ROBOTS MAKE GOOD SONGS
It’s:
WHAT THE HELL IS HAPPENING TO THE MARKETPLACE
QOBUZ HAS ALSO BEEN CLEANING THE SEARCH RESULTS
The company says it has removed from general search more than one-third of the AI-generated albums it identified that had no listening activity, representing roughly four million tracks. Those releases can still be found when users search specifically for their titles, but Qobuz is deliberately keeping them from flooding normal discovery.
Four.
Million.
Tracks.
Think about that.
There was a time when getting a song recorded required musicians.
Studio time.
Production.
Engineering.
Mixing.
Mastering.
Distribution.
Promotion.
Possibly six people arguing about the snare drum for four days.
😂
Now enormous volumes of audio can be created at machine speed.
That does not automatically make the music worthless.
But it changes the economics of abundance dramatically.
AND THIS IS WHERE HUMAN CURATION BECOMES MORE IMPORTANT NOT LESS
Qobuz says its featured playlists, albums of the week, editorial selections, reviews and other major discovery surfaces remain selected by human editorial teams. The company also says the AI-generated material it identifies is kept out of those human-selected discovery areas and out of its algorithmic recommendation system.
That is fascinating.
Because while everyone keeps telling us artificial intelligence will make human music programmers obsolete, Qobuz appears to be reaching almost the opposite conclusion.
The more content machines can produce…
THE MORE VALUABLE A TRUSTED HUMAN FILTER MAY BECOME.
Hello, radio.
Are you paying attention yet?
BECAUSE THIS IS A RADIO STORY TOO
Absolutely.
If an ocean of music is coming…
Who chooses what matters?
That has been radio’s job for approximately forever.
People act like music programming is simply deciding which 300 records belong in a computer.
No.
Great music programming is trust.
Somebody says:
This record matters.
This artist matters.
This deserves three minutes of your life.
Listen.
That judgment becomes increasingly valuable when the supply of available music becomes effectively infinite.
THE MACHINE CAN MAKE A MILLION SONGS
Wonderful.
Which five should I hear?
There’s your business.
😂
And that is why I believe radio should be watching Qobuz closely.
Not because radio should automatically copy every rule.
It shouldn’t.
But the philosophy is strong.
AI CAN ASSIST
HUMANS DECIDE
THE AUDIENCE GETS TRANSPARENCY
That’s remarkably close to where I believe broadcasting should land.
I AM NOT AFRAID OF AI MUSIC
I want to make that perfectly clear.
I’ve already said this repeatedly.
Artificial intelligence is no longer the future.
It is part of the present.
People are going to make music with it.
People are going to make commercials with it.
People are going to make imaging with it.
People are going to use it to write.
Produce.
Edit.
Research.
Translate.
Generate.
Experiment.
Some of that work will be extraordinary.
Some of it will sound like two refrigerators having sex inside an elevator.
😂
That is also true of human music.
The existence of bad AI music does not make AI inherently bad.
THE PROBLEM IS DECEPTION
That’s where my line gets considerably darker.
If the audience believes something is human when you know it is synthetic?
Tell them.
If a voice was cloned?
Tell them.
If an entire artist does not actually exist?
Tell them.
If the song was substantially generated through AI?
Tell them.
You don’t have to stop people from listening.
JUST STOP PLAYING GAMES WITH WHAT THEY THINK THEY ARE HEARING.
That is what Qobuz gets right.
AND QOBUZ ITSELF USES AI
This may be my favorite part because it destroys the simplistic argument immediately.
Qobuz isn’t anti-AI.
Its September charter explicitly says the company uses AI in areas such as transcription, moderation, first-level customer support, search improvements, security and repetitive operational tasks, while requiring human validation of AI-assisted work.
There you go.
AI detection is itself being used to police AI music.
😂
Welcome to 2026.
The technology isn’t evil.
The question is what are you using it for.
QOBUZ DRAWS THE LINE AT REPLACING THE HUMAN EDITOR
And that is where radio needs to have a serious conversation.
Qobuz says it will not replace editorial judgment with AI.
Its people still select its major playlists.
Its people still write its reviews.
Its people still decide what receives editorial prominence.
Why?
Because taste is part of the product.
Radio programmers should frame that sentence and put it on the office wall.
TASTE IS PART OF THE PRODUCT.
WE HAVE SPENT YEARS TRYING TO REMOVE TASTE FROM RADIO
Research it.
Test it.
Score it.
Rank it.
Rotate it.
Analyze it.
Benchmark it.
Normalize it.
Then somebody wonders why everything starts sounding remarkably similar.
Data matters.
Research matters.
Algorithms can be incredibly useful.
But somebody still has to decide:
THIS RECORD BELONGS ON THIS RADIO STATION
That is not simply mathematics.
It is programming.
NOW IMAGINE WHAT HAPPENS WHEN MUSIC SUPPLY BECOMES UNLIMITED
This is where everything changes.
Historically there were practical limits to how much commercially viable music could enter the marketplace.
Making records cost money.
Distributing them cost money.
Promoting them cost money.
Human time itself created scarcity.
Generative AI destroys much of that scarcity.
A system can generate music continuously.
Twenty-four hours a day.
Seven days a week.
It doesn’t need lunch.
It doesn’t need sleep.
It doesn’t need a hotel room.
It does not get into a screaming argument with the bass player in Des Moines.
😂
Now imagine thousands of systems doing that.
The problem becomes less:
CAN WE CREATE ENOUGH CONTENT
and more:
HOW DO WE FIND THE CONTENT WORTH OUR TIME
That is curation.
AND RADIO WAS BUILT FOR CURATION
This is the opportunity.
Instead of fearing AI music, radio can become the trusted gatekeeper that explains it.
Here is an AI-assisted song.
Here is a fully synthetic record.
Here is a human artist using generative tools.
Here is why this song is interesting.
Here is why people are angry about it.
Here is the artist responding.
Here is the label explaining it.
Here is the songwriter talking about it.
MAKE CONTENT OUT OF THE QUESTION.
Radio has microphones.
Use them.
I WOULD EVEN GO FARTHER THAN QOBUZ
If I programmed a station and chose to play a fully AI-generated record that became culturally significant, I would not bury that information.
I’d promote it.
COMING UP THE RECORD EVERYBODY IS FIGHTING ABOUT BECAUSE THERE IS NO HUMAN SINGER
Now you’ve got appointment listening.
😂
Then take calls.
Do listeners care?
Can they tell?
Do they like it?
Would knowing change their opinion?
That’s radio.
We should not fear interesting things.
BUT VERIFY WHAT YOU ARE LABELING
This matters too.
Qobuz openly acknowledges that AI detection is not perfect and says its systems will continue evolving as technology changes. It provides a process for artists who believe material has been incorrectly labeled and a reporting route for suspected AI material that escaped detection.
GOOD.
Because falsely identifying somebody’s human work as machine-generated creates its own problems.
Transparency has to include humility.
If you don’t know…
say you don’t know.
THE INDUSTRY NEEDS A COMMON LANGUAGE
AI-generated.
AI-assisted.
Synthetic voice.
Human performance.
Generated instrumentation.
Cloned performance.
Prompt-created track.
There are enormous differences among those things.
Throwing all of them into a bucket labeled AI MUSIC is intellectually lazy.
A producer using AI noise removal is not the same thing as generating an entire singer.
A songwriter brainstorming with a tool isn’t necessarily the same as pressing a button and accepting a completed song.
Context matters.
Qobuz’s current system targets material it identifies as AI-generated rather than pretending every technological contribution deserves the same treatment.
That’s another good lesson.
ON THE DIAL PERSPECTIVE
I think Qobuz is onto something.
Not because I agree with every possible implication of its policy.
But because the company has chosen something far healthier than panic:
KNOW WHAT IT IS
LABEL WHAT IT IS
LET HUMANS CURATE
LET LISTENERS DECIDE
That works for me.
THE BEST AI POLICY MAY NOT BE NO
It may be:
TELL ME
Tell me how it was created.
Tell me whether the singer exists.
Tell me whether somebody’s voice was cloned.
Tell me whether humans wrote it.
Tell me whether AI assisted.
Then let me experience the work.
If I love the song, I love the song.
If I hate it, I hate it.
At least I know what I’m evaluating.
AND HERE IS WHY RADIO SHOULD BE PAYING CLOSE ATTENTION
Qobuz says AI-identified music represents only 0.38 percent of listening on its platform.
Yet the company has already encountered enough synthetic catalog volume and suspicious streaming behavior to build proprietary detection technology, remove millions of tracks from ordinary search exposure and rewrite its AI Charter within months of issuing the first version.
That tells me the music business is moving quickly.
Very quickly.
Radio does not have five years to start thinking about this.
The conversation is happening now.
QOBUZ JUST DREW A LINE
Not between technology and humanity.
Between technology and deception.
Between automation and judgment.
Between infinite supply and trusted curation.
That’s a much smarter line.
And maybe radio should draw one too.
Play the AI record if it deserves to be played.
Use AI behind the scenes if it makes the station better.
Experiment.
Learn.
Innovate.
But when the technology materially changes what the audience believes it is hearing?
TELL THE DAMN LISTENER.
Because the future of music may contain humans.
Machines.
And increasingly some combination of both.
The one thing it should not require is a detective.
#OnTheDial #Qobuz #AIMusic #ArtificialIntelligence #MusicStreaming #Radio #RadioProgramming #MusicRadio #GenerativeAI #HumanCreativity #MusicIndustry #StreamingMusic #RadioIndustry #Broadcasting #MusicDiscovery #FutureOfRadio