I first noticed it a couple of months ago, on one of those ambient/LoFi playlists that Youtube sends you to randomly. It wasn’t bad, but had a few weird bits. New account, artist I’d never heard of. When I checked the description, there it was: “AI music”.
Turns out, this has become a common occurrence.
It’s hard to say just how much AI music there is on Youtube, but you can find a bunch of channels that pour out hours and hours of “new” music every week. It’s usually the same, generic, thematic instrumental music. But not always.
Scroll through YouTube or Spotify, and you’ll soon stumble upon names like Concubanas and Phantasia and you’ll find flawless blends of salsa, rumba, and Congolese jazz.
“This multicultural ensemble was formed in Havana in 1971, creating a unique fusion of Cuban and Congolese music. The group’s name would prove prophetic, as they became one of the most significant examples of artistic cohabitation between African and Latin American musical styles of their era,” writes the description on one Concubanas album.
Except none of this is real. The band doesn’t exist, and it’s all AI.
Your new favorite band might not be real
I’ll be honest, Rumba Congo (the fake album by the “band” Concubanas) doesn’t even sound bad. You can find some minor issues, but even listening to it and paying attention, it sounds okay. Phantasia is another example, a band that “entered a period of silence” in 1976. These are finessed, pleasant sounding products. You’ve probably heard worse man-made music.
Turns out, AI music is already big business. A North Carolina man used AI to create hundreds of thousands of fake songs and fake bands. He put them on various streaming services and collected a very non-artificial $10 million. At least that’s what prosecutors said when they charged him with fraud.
The tracks, which had names like “Zygotic Washstands,” “Calvinistic Dust” and “Zygophyllum,” became top performers on Amazon Music, Apple Music and Spotify, reaping millions and millions of streams.
But this is just one guy. The International Confederation of Societies of Authors and Composers (CISAC) predicts AI music revenue will soar from $100 million in 2023 to a staggering $4 billion by 2028, making up nearly a fifth of streaming income. Let’s put it this way: AI will be worth 20% of the streaming music market.
Unsurprisingly, most listeners have no idea if what they’re listening to is AI-made.
No AI beats to relax and study to
If you’ve streamed a lot of music, the odds are you’ve come across some version of “LoFi Girl.” In particular, one channel offers endless loops of mellow beats from lo-fi and chill-hop genres, paired with a distinctive animated girl who is usually reading or studying. Dozens of imitators now stream similar visuals and indistinguishable music.
Some of them are undoubtedly AI. But according to TikTok music influencer Derrick Gee, even the original channel (which has 15 million subscribers) is almost entirely AI-generated. Gee provides convincing evidence and there’s growing evidence that AI is slowly but surely infiltrating many of our favorite playlists.
Perhaps the worst part about this is that the streaming companies seem to be okay with this.
YouTube requires creators to note “altered or synthetic” content, but it’s often buried deep in descriptions. The label may not appear properly on some devices. Violations could prompt YouTube to apply its own labels or even remove content, but it’s hard to quantify how well this is enforced. Meanwhile, Spotify has no clear labeling policy. Co-president Gustav Söderström sees AI as a creative enabler and says AI music is welcome — while worrying only about legal pitfalls like copyright. There’s no talk of disclosure or labels.
Unsurprisingly, users aren’t thrilled by the influx of AI music. Many users feel betrayed, with one Reddit user rightfully pointing out that real artists are going to “have an even harder time getting discovered because people will be so cognitively stunted by AI.”
But the vast majority of users simply aren’t aware. What feels like just another song may in fact be a ploy to make money. And perhaps, we shouldn’t even be surprised.
AI is affecting all art
AI in music mirrors AI advancements in other creative fields like writing and visual arts. Just as deepfakes challenge trust online, AI-generated music now tests our ability to connect with creators and understand authenticity.
Of course, if applied thoughtfully and ethically, AI can empower musicians — offering tools for co-creation, remixing, and ideation. Research involving real artists shows promise for human-AI collaboration.
Without transparency and ethical standards, listeners are deceived and artists are once again overshadowed. As for the AI music creators, they’re just getting started. Concubanas and Phantasia are both creations of a relatively small YouTube channel that is just nine months old and has “only” around 40,000 subscribers. Yet, in that short time, it has published over a dozens of elaborate musical fictions, each with its own invented band, backstory, and sonic identity. The channel includes a disclaimer about “altered or synthetic content,” and on its main page, it even offers a philosophical shrug disguised as a provocation:
“You’re not going to believe your ears! Everything that happens on this channel is fiction. But what is the truth? F*ck it, just listen!”
After all, does it really matter if the music we listen to has any humans in it? That’s a question we may need to answer sooner rather than later.
