Deslopification: Why AI Music Needs Its Own Room
Bass Battle 2025 hosted by DnB Seoul
I want to be upfront about something before I say anything else: this isn't an anti-AI post. I'm not interested in pretending the technology doesn't exist, or that it isn't going to keep getting better, or that everyone using it is acting in bad faith. Some of it is genuinely impressive. Some of it is being used by real musicians as a real tool, the way a drum machine or a pitch corrector is a tool.
But somewhere in the last year, I stopped being able to open a "Discover" playlist without wondering whether the song I was hearing had ever been felt by anyone before it reached me. And I don't think I'm alone in that.
The flood is real, and it's not slowing down
By June 2026, more than half of everything uploaded to Deezer on a given day, over 90,000 tracks, were AI-generated. That's up from 30% in December 2025 and 44% in April. At that rate, some analysts project the platform could approach full AI saturation by late 2027. Apple Music has reported that more than a third of its uploads are now 100% AI-made. This isn't a niche corner of the internet anymore. It's the mainstream pipeline.
A lot of it isn't even there to be listened to. Deezer has been explicit that much of the flood is fraud: bulk-upload AI tracks, farm them for streams, often bot streams, and skim a slice of the same royalty pool that real musicians are paid from. Apple demonetized two billion fraudulent streams in 2025 alone. That's not a debate about creativity. That's theft, wearing music as a costume.
Where the industry is starting to draw lines
The interesting part isn't that platforms are worried, it's how they're starting to respond, and it tells you something. Tidal now puts an "AI" badge on fully AI-generated tracks and cuts them out of royalty eligibility entirely. Apple Music rolled out transparency tags across audio, composition, artwork, and video, so listeners can actually see what they're getting. Deezer's detection system pulls AI tracks out of recommendations and playlists rather than letting them compete head-to-head with human work for the same algorithmic real estate. Even SoundCloud, which has stayed the most permissive, had to introduce opt-in consent after backlash over AI models training on artists' work without asking.
None of these platforms banned AI music. What they did was draw a boundary, a badge, a tag, a separate lane, so that AI-made and human-made music aren't competing for the same trust, the same discovery slot, the same royalty pool, without anyone knowing the difference. That's the part I think is actually right, and it's the part I want Harmonize to stand for too.
Listeners already know the difference and they're telling us
Here's what convinces me this isn't just a founder's pet opinion: the data says listeners feel it too. Luminate's research found that comfort with AI-made music dropped from -13% to -20% between May and November 2025, and it's getting more negative, not less, especially among Gen Z and Gen Alpha, the generations everyone assumes are supposed to be the most AI-native. The most disliked use case of all is AI mimicking a real artist's voice or style; the exact thing that erases the person behind the music entirely.
And despite AI tracks making up close to half of daily uploads on some platforms, they account for less than 3% of actual human streams. People are, without being told to, voting with their ears.
Borrowing the word ‘deslopification’ from a different corner of the art world
There's a term that's been circulating in AI art circles for a while now: deslopification; the process of cleaning up, humanizing, and restructuring messy AI outputs into something that actually respects the people encountering it. It hasn't really crossed over into conversations about music yet. I think it needs to.
Applied here, deslopification isn't "get rid of AI." It's "stop letting AI-made and human-made music share a room without anyone being told which is which." Music has always been how we've processed grief, marked a summer, said something to a friend we couldn't say out loud ourselves. That only works because there's a person on the other end of it, someone who felt something and decided to put it into a melody so someone else could feel it too. An AI model doesn't grieve. It doesn't remember the person a song was written about. It can imitate the shape of emotion, but it didn't earn it, and it isn't in community with the people listening to it the way a human artist is.
That's not an argument that AI has no place. It's an argument that its place shouldn't be indistinguishable from a human's place; not on a chart, not in a recommendation feed, not in a "you might also like" slot next to a song someone's grandmother used to sing.
Where Harmonize stands
We built Harmonize because we believe music discovery should run through real people; the friend with impeccable taste, the curator who's been building a scene for a decade, the community that shows up for an artist because they mean something to them. That belief doesn't require us to be against AI as a technology. It requires us to be for clarity about what's human and what isn't, and for keeping the two in separate lanes rather than letting one quietly absorb the other.
So consider this a starting point, not a finished policy. We're going to keep building Harmonize as a place where the music you find came from somewhere you can actually trace back to a person. If that's a boundary the rest of the industry is only just beginning to draw, we're glad to be drawing it early.
What do you think the boundary should look like? I'd genuinely like to know. Let us know on Threads, Instagram, Facebook or LinkedIn!
