Can You Actually Hear the Difference Between AI and Human Music?
- 18. Juli
- 4 Min. Lesezeit
The Human Frequency — Notes on making music by hand in the age of AI
By Roger Ricks
I want to start with a number that honestly unsettled me a little. In a survey of around nine thousand people across eight countries, ninety-seven percent of them could not reliably tell fully AI-generated music from human-made music in a blind test. Ninety-seven percent. So if you were quietly counting on your own ears to keep you safe, well, I was too, and that number put me right back in my place.
The night a machine gave me goosebumps
I should say something up front so we understand each other. I don't make music to pay my bills. I earn my living in real estate development, and the music is the thing I do because I honestly can't not do it, with every stream of it going to charity. So I've got no reason to oversell any of this, and no reason to pretend AI music is worse than it really is. And I'll be honest with you about a moment that genuinely threw me. Not long ago I sat and watched Rick Beato build a track from almost nothing in a few minutes, a song that went out under the name Sadie Winters, and I watched the whole thing with this strange mix of fascination and unease. Because the song actually got to me. It carried real emotion. When the chorus hit I got goosebumps, honest to God goosebumps, and I sat there a little stunned that a thing made that fast, that way, had just reached into my chest and pulled on something. I did not expect that, and I am not going to pretend it didn't happen.
What I still listen for
So is the difference already gone? Not completely, and not yet, but moments like that one are why I can't be smug about it. Let me tell you what I still listen for, and maybe you'll start to hear it too. The female voices have become genuinely convincing, smooth and expressive in a way that is hard to argue with. It is the male voices where I usually still catch it. They tend to land in the same place, the same character over and over again, as if they all grew up in the same house and never left it. Play four or five AI tracks back to back and after a while you notice it is basically one voice trying on different jackets. Then there are the small artifacts, a strange little shimmer at the front of a word, a note that smears where a real singer would just take a breath. Once you have heard a few of them you can't really unhear them again. But those tells are getting rarer by the month, and the Sadie Winters thing was proof of that to my own ears.
Forget the first listen. Trust the tenth
So where does that leave me, a guy who got goosebumps off a machine and still wants to stand here and tell you there is a difference? Here it is, as plainly as I can manage. I don't trust the first listen anymore. I trust the tenth. This stuff is engineered to win you over the moment it starts, all polish and confidence, and yes, it can clearly land a punch, the Sadie Winters chorus is proof of that. What actually keeps me up at night is a different question. What is left of that song a week later? A year later? Will I be reaching for it in ten years the way I still reach for the records that have somebody's whole life bleeding through them? I really don't think so. The goosebumps were real and I won't take that back, but the longer I sat with it afterward, the more it felt like a beautiful room with nobody living in it. Gorgeous, and completely empty. And I have started to believe that the emptiness is the part you keep feeling long after you have forgotten whether your ears could even tell.
Try this yourself this week
So try something this week. The next time a song actually stops you, the next time one pulls you back in, ask yourself whether you would care if a machine had made it, and whether it would still hold up on the tenth time round, and the hundredth. I have a feeling your answer will tell you far more about what music is really for than any survey ever could.
One last thing before you go
And here is the thought I'll leave you sitting with. There is one sound in my own music that could never come from an AI, not the Sadie Winters kind, not any of them. Technically it is a flaw, and for years I was a little embarrassed by it, and now it sits on every single track I have ever recorded, and it might just be the most honest thing about me. Next time I am going to play you my new single, "That Day," and show you exactly where that flaw lives and why I would never in a hundred years fix it. So come back for it. Or if you really can't wait, it is already up here, and you are welcome to go looking for it yourself before I point it out. So tell me honestly, has an AI song ever reached into you the way Sadie Winters reached into me? And if one ever did, did it stay with you, or was it gone by the next morning? I read every single comment.
Sources & further reading (where they actually tested this)
Deezer / Ipsos listener survey (2025) — around 9,000 people in eight countries; 97% could not reliably tell fully AI-generated music from human-made music in a blind test. https://newsroom-deezer.com/2025/11/deezer-ipsos-survey-ai-music/
Wu & Holmes, Cognitive Research: Principles and Implications (2026) — a peer-reviewed study: labelling a piece "composed by AI" led listeners to imagine fewer and less vivid stories, even when a human actually made it. https://link.springer.com/article/10.1186/s41235-026-00715-z
Ansani et al., Music & Science (2025) — "AI performer bias": people rated the very same performance lower when told it was played by an AI. https://journals.sagepub.com/doi/10.1177/02762374241308807
Friedrichsen, Schwarz & Clement (2026) — listeners' willingness to pay for a track dropped once its AI origin was disclosed (working paper). https://www.promarket.org/2026/05/04/consumers-prefer-ai-music-until-theyre-told-its-ai/


Kommentare