Considering most music uses sampling I doubt this is even close to possible.

It's not really possible, especially given that most music is touched up post production to begin with (so isn't "natural" anyways).

Much like detecting LLMs or AI photos/videos it's mostly heuristics and so it can only give you a probability something was AI generated. Even worse, any improvement on detection can be used to improve the models that make the content.

You can always check RYM to see how other people are rating/reviewing the artist.

What useful information are you needing here that you don't get from listening?

Is your decision to listen to a song influenced by an AI purity test?

"This sounds like derivative crap but it's not AI so I guess I'll listen to it"?

"I enjoy this song! But it has 62% AI so no actually I hate it now"?

I mean, there are apps that can detect Auto-Tune, I expect there will be apps to detect AI. Maybe not YET, but eventually.

I don't think it's easy to do. Given how unreliable "AI detectors" are in general.

Also, why? Music is something very sensitive driven. If you like it you like if you don't you don't, I don't think a quantitative measure on how a song is made is a reasonable approach to distinguish which songs you like and which song you don't.

I can just imagine:

• Do you like this song?

• I don't know yet. (Pulls phone out to measure AIness of the song) No I don't like it.