If No One Wants AI Music, Why Does It Feel Like It’s Being Forced On Us?

In my recent experience speaking to colleagues, artists, bands and just anyone who listens to music, no one seems to like, want, need, desire or knowingly listen to AI-generated music. I would say that my overall takeaway from these conversations has been that people are quite dismissive of it entirely.

Taking that in mind, something quite odd happened in the music business world this week. Between Tuesday and Thursday, four companies made four separate moves on AI-generated music, and none of them are moving in the same direction.

BMG signed a global licensing deal with Suno on Wednesday. Yes, that Suno, one of the two AI music generators the major labels sued back in 2024 for training on their recordings. And, this new deal settles any of Suno’s previous use of BMG’s catalogue. A day earlier, Spotify had announced AI Persona badges, so, from mid September 2026, profiles built around artist identities which have been generated by software will carry a visible label. And, importantly, their music will be excluded from editorial and algorithmic recommendations by default. Beatport went the other way entirely, expanding its partnership with Beatdapp to detect AI tracks across its stores. And then there was D’Addario, a company that sells strings to guitarists, admitting at last that the music in its NYXL promo was AI after weeks of avoidance.

So, in the space of roughly 48 hours, a major rights company has paid for a seat at the table while a streaming service decided the safest place for this music is outside its recommendation algorithms. A dance retailer has put money into catching it before it hits the shelves. And a string manufacturer apologised for even using it at all. What’s going on here? Either somebody in that list has misread the market, or nobody has, which would be far stranger, and that is where this piece will be going.

All of the above raises two quite simple questions for me. If the people who actually listen to music keep saying no to this AI generated music, who exactly is it being made for? And, who are these deals for?

About the Author

Ron Pye is the CEO and Managing Director of IQ Artist Management, a music management and consultancy company based in the UK. They work with artists on a global scale. Ron was awarded an MA in Music Industry Studies from the University of Liverpool, with Distinction in 2024, and a BA in Music Business and Finance from the University of Middlesex.

Streaming royalties are a working part of his week. Statements, licensing terms and platform policy changes cross his desk as documents that decide what artists get paid rather than as industry news. When a platform rewrites its rules, the practical question is what the change does to a client’s next statement, and the money behind those statements comes from listener subscriptions, so what platforms let onto their services is something he watches closely and professionally.

His writing on AI music comes from inside those payment systems, as the person answerable to artists for what actually arrives on the statement.

What Listeners Keep Telling Anyone Who Asks

Let’s start with the listener appetite I have just alluded to. Luminate, the company that surveys and supplies the industry’s consumption data, measured ‘listener sentiment’ (how they feel about AI being used in music) twice in 2025. The net figure fell from minus 13% in May to minus 20% by November, with the steepest drop among Gen Z and Gen Alpha, the very listeners every platform is chasing hardest. So, the user position on AI Gen is already negative amongst the users who use streaming platforms practically every day, and, it’s getting worse.

User behaviour is also backing that data up. Deezer recently published that more than half of all daily uploads to its platform are now fully AI. That’s around 90,000 tracks a day, yet fully AI music only makes up between 1% and 3% of what people there actually listen to. And, 85% of those AI streams are fraudulent, so, bots are streaming machine made tracks to farm royalties, the same fake engagement problem that has changed what streaming numbers even mean. If we do the maths on those two figures. AI tracks get 3% of streams at most, and 85% of those streams are bots.

So, half of all of the declared daily uploads, are listened to by almost nobody.

A small caveat: Deezer sells AI detection tools these days, so it has a commercial interest in the problem looking big. It is also the only platform publishing these figures at all, Spotify, Apple and Amazon release nothing comparable, so there is nothing to check Deezer’s numbers against. What there is, though, is how everyone else is behaving. Spotify is badging, Beatport is paying for detection, and TIDAL has cut the money off entirely. Nobody buys detection tools unless there is something to detect.

When Deezer and Ipsos surveyed 9,000 people across eight countries, 97% could not pick the AI tracks from the human ones in a blind test. And, roughly a third of listeners said they did not much care either way. So, this isn’t a story about everyone hating AI music, a decent chunk of the public is indifferent, and most of us cannot reliably hear the difference anyway.

Interestingly though, what the survey did find was overwhelming support for labelling fully AI music. A label is there so you know what you are getting, and people only demand one when the difference matters to them.

D’Addario is the consumer market enforcing this point. A string maker’s whole customer base is musicians. They used AI music in a promo, then spent weeks denying it, then they folded in public: “We do not support AI-generated music”. Then, they went further and spelled out why: “If AI replaces musicians, we’re out of business”. I don’t need to explain that companies do not issue apologies for giving customers what they want.

Where Each Platform Currently Stands

By now every service selling music to the public has picked a position on AI, and the positions run from outright bans to building AI tools of their own.

In January, Bandcamp banned the stuff entirely under the banner “Keeping Bandcamp Human”, the plainest policy statement any platform has made so far. Amazon sits at the other end: no available public AI music policy at all, and it put a Suno integration into Alexa+ while the majors, including its own licensing partners, were suing Suno. One industry source told Billboard it was “pretty ridiculous” for the owner of “a licensed, legitimate music service” to ship a product like that in the middle of the lawsuit. Or, was it? TIDAL keeps fully AI tracks on the platform but stopped paying them anything at all in July. Apple now requires transparency tags. Qobuz wrote a charter banning full AI submissions from its catalogue. And Spotify has managed to take two positions at once, badging AI personas on the listener side while licensing AI remix technology with UMG and Merlin on the business side.

The full picture, as it stands, looks like this.

Last reviewed: 14 August 2026

A first glance suggests that the smaller platforms are strict while the giants are not, and there could be something in that. That’s being said, Deezer is hardly a minnow and sits in the strict camp. An outlier, maybe? But then Spotify, the biggest music streaming service in the world, manages to be in both camps at once. I’d wager two exceptions stops being a coincidence. Something other than size is deciding who bans this stuff and who builds it.

The Closer You Are to the Listener, the More You Say No AI

What separates the top of that table from the bottom is not company size alone, it’s how ‘close’ the app feels to the person doing the listening.

There’s Many Music Apps That Feel Nothing Alike

On your phone, Qobuz and Amazon Music look pretty much like the same thing: a music app, a search bar, a play button. All the sharp UI and UX you’d expect. Using them, though, makes you feel different. Qobuz feels like something closer to a community, a service run by people who care about the same things you do. There’s a ‘sticking it to the man’ kind of vibe going on, even though that isn’t happening. Amazon Music feels like being a customer number inside something absolutely enormous. And, behind the apps, they are very different, in terms of what music really means to them and in sheer scale.

Qobuz sells music subscriptions and nothing else. They are fighting for every single user they can, so if a music fan decides to walk away and airs their view online, the company would feel that at the highest levels. It is small enough that its customers’ opinions actually reach the upper echelons, and it behaves like a company that knows it. In contrast, Amazon Music is one corner of an absolutely gigantic shop that sells everything, and is owned by one of the biggest companies on Earth. If the music app closed tomorrow, Amazon would barely notice. Bezos would get a bloody nose, but he’d move on with the next thing.

Apple and YouTube sit in the same position: music apps out front, a phone maker and an advertising business behind them, both companies so vast that music could vanish from them entirely and I’d bet the share price would hardly move. That is what ‘distance from the listener’ means to me, and what it means in this context. Partly it is what the company stands to lose if people who love music stop choosing them. The other part is the sheer size, because scale pushes the listener further away from the decision makers: a complaint lands on a desk at Qobuz, while at Amazon it disappears into a system. The more a company has riding on music fans, and the nearer those fans sit to the people making decisions, the harder it fights against AI music.

Running the Test Down the Table

Bandcamp’s customers are fans paying musicians directly, and it has banned the music outright. TIDAL and Deezer sell music subscriptions and nothing else, and between them they defund it, tag it, and keep it out of recommendations. Beatport’s trade is DJs buying tracks to play to a crowd, and it pays for detection. At the other end, YouTube embraces the technology and builds its own, with Neal Mohan assuring the industry through Lyor Cohen’s newsletter that “AI will remain a tool for expression, not a replacement”. SoundCloud earns most of its keep from uploaders, and asks for disclosure about AI and little else.

Notice what changes along that list. The same AI track Bandcamp bans is perfectly welcome on YouTube. So, whatever drives these decisions, it can’t be the music. The only thing that changes is the distance felt between the company and a human being who presses play.

Spotify: The Exception That Proves It

Spotify looks like the exception and turns out to be the strongest confirmation. One side of its business is listener subscriptions. The other is label relationships, a marketplace, and advertising. Sat in the middle of the table, it behaves exactly in the middle: AI Persona badges and default exclusion on the listener side, shutting that music out of the playlists and recommendations where the majority of streams happen, while the rights side licenses AI remix technology with UMG and Merlin. Its own announcement leans on the phrase “authentic artists”, which takes some front from the platform whose open uploads helped flood streaming with the stuff in the first place. I made a similar point about Spotify’s Verified badge in May: the badge protects the listener experience, because the listener experience is the product.

But two of the companies at the centre of this week’s news are not in that table at all. Record labels make their money by selling permissions to other companies and they have now sold those permissions to the AI firms. Suno’s money comes from the people making tracks with it, not from anyone listening. The companies that just signed this massive deal do not have listeners anywhere near their business models.

Sue First, License Later (Maybe)

In June 2024 the three major labels, Universal, Sony and Warner, sued Suno and Udio through the RIAA, the American record industry’s trade body. The accusation was simple: both companies had trained their AI models on the labels’ recordings without asking anyone, “on an almost unimaginable scale“ as the filing put it.

Then, one by one, the labels stopped suing and started signing. Sixteen months in, Universal settled with Udio. In practice that meant Udio paid for what it had already done, agreed licence terms for the future, and the two are now building a licensed AI music platform together, due later this year. Warner followed a month later, settling with both companies: artists’ names, voices and songs get used only if the artist agrees. And then last Wednesday BMG, the fourth biggest music company in the world, joined in without ever having sued anyone, going straight to a licensing deal with Suno that also settles any claims over past use of its catalogue.

What the BMG Deal Confirms

We don’t know what BMG has been paid as no figure has been published and neither company has explained how the money will flow. The confirmed terms amount to three things: BMG artists and songwriters are included only if they choose to be, those who take part get paid, and the fight over Suno’s past use of BMG’s music is now over. How much, how often, and on what terms we may never know.

If we take the Warner and Universal deal as the template, the likely shape is a licence fee up front plus a cut of the subscription fees Suno collects from its users.

Sony is still the odd one out. It remains in court with both companies as of the most recent reporting. Universal sits in the strangest position of all: still suing Suno while in business with Udio two companies accused of exactly the same thing. If this were about the principle of training AI on artists’ work without permission, you could not sue one company over it and take money from the other, could you? I feel somewhere along the way, these lawsuits stopped being about whether AI training was wrong and became more about the price of making it legal.

Forbes spotted this back in December and gave the pattern a slightly cynical name that stuck with me: launch, train, settle. Build the product first. Train it on music you never paid for. When the lawsuits arrive, settle them, and walk away holding a licensing deal.

And I, and many others, will have noticed who was missing from every one of those negotiations: the listeners. These deals were signed between rights companies and technology companies. What changed hands was permissions, with catalogue access flowing one way and money flowing the other. A normal music deal is a bet on an audience: a label signs an artist because it believes people will pay to hear them. Whether anyone ever listens to a single AI track has no bearing on what Suno owes the labels, so the question never came up in two years of suing, settling and licensing.

Two Reasons to Take the Money

Why did the companies settle though? Why not take Suno and Udio to the cleaners? Nobody involved has said, but two explanations come through the overcast skies in my mind. The first is that the labels might have lost. Whether training an AI on copyrighted recordings counts as fair use was untested for music. In June 2025, while these lawsuits were live, two US federal judges ruled in the same week that training AI models on copyrighted books was fair use. The applied logig being that the AI was not copying the books, it was learning from them, the way a person reads a pile of books and then writes something of their own. Stealing (pirating) the books in the first place was a different matter. Universal settled four months later. Settling converted a legal risk into a predictable income stream.

The second explanation is far simpler. It turns out that the AI companies are worth more to the labels alive than dead. Money enters this system at the end where music gets made. At the end where music gets heard, the platforms are banning it, demonetising it and keeping it out of recommendations. How a business can run like that, money flowing in one end while the product gets turned away at the other, is the final piece of the puzzle.

An Engine That Runs Without an Audience

So, let’s bring this full circle and answer who is this music being made for, and who are these deals really for? The answer is sitting in Suno’s own accounts.

Suno’s own figures show it has 2 million paid subscribers and $300 million in annual recurring revenue. In June it raised $400 million after a $5.4 billion valuation. And that pot is where the major labels’ new licensing money comes from. When BMG or Warner take their cut, whatever those cuts turn out to be, it comes out of the $300 million a year that people pay Suno to make tracks.

Which leads to a position I find genuinely strange. The royalty that will eventually reach an artist who opts in to AI training will be funded by other people making AI music. Hobbyists paying their monthly subscriptions, not by anyone choosing to listen to anything. The whole royalty chain, from Suno subscriber to label to songwriter, is music makers paying music makers, with a technology company keeping its margin in the middle.

It also starts to answer the question this article opened with: if nobody wants AI music, why do 90,000 AI tracks a day keep arriving on Deezer? Because there are two economies running here, and neither one needs a listener.

The first one we have already met. Suno’s two million subscribers are, in the main, people paying to make music for the pleasure of making it, rather like a gym membership that never quite gets used. You maybe turn up a few times a month, just to make yourself feel a bit better. Their subscriptions are the $300 million, and whether anyone ever hears the results was never the point.

The flood comes from the second economy. A study published in June did the sums: a top Suno subscription can churn out 6,000 tracks a year at under two cents each, and getting all of them onto streaming costs about $26 a year. Most earn nothing. Only 7% of the AI tracks on Spotify picked up enough plays to be paid at all, and three quarters of those made less than $10. For the uploaders the maths still works, though, because the tracks cost almost nothing to make. It’s simple spam economics. Flood the platforms and let the rare winner pay for the rest. And there are winners. Kapwing estimates that Breaking Rust, the AI country act, has earned $469,963 across Spotify and YouTube.

As for who is doing the “listening,” the same study describes the technique. Stream each track a few thousand times, enough to earn, low enough to slide under fraud detection built for high volume replays. Deezer’s 85% figure suggests how much of it is machines making the tracks and the machines listening to them.

If we put all of the pieces together, the listener surveys, the platform rulebooks, the settlement terms and the economics study, they all come from different corners of this industry. Those corners barely speak to each other and have no reason to agree. Yet they all seem to point in the same direction.

What All of This Adds Up To

Two things are also true at once. As a business, AI music clearly works. Subscription money comes in every month, the labels now take a cut of it, and serious investors have valued Suno in the billions. As music, it has failed the one test that counts the most, hardly anyone wants to hear it. Weirdly, the business runs along quite happily with both of those being true.

This business now makes its money whether anyone is listening or not. AI music is not made for listeners. It is made by hobbyists who pay Suno for the fun of it, and by spammers chasing royalties. The labels will now get paid from those subscriptions either way. Nobody in that chain needs a single person to press play.

So, if you make music for people who want to hear it, I’d argue you are in a different business from the one described in this article.