I would have thought similarly, but actually feeding 19th century poems to Suno and iterating on the prompts several times I got some results that moved me emotionally, as in, listening/reading the words with this musical presentation enhanced my appreciation of the poems and it felt more visceral. Like making angry revolutionary poems into grunge brought it closer and less of a "histoic", "bookish", "dusty" thing.
I think there is a great case to be made here using purely synthetic sounds as the basis for emotion. Vangelis (Soil festivities), Klaus Doldinger (Skyscape) are great examples. These are sounds that have been produced exclusively by the mind and in spite of there not being a physical instrument involved they manage to convey imagery and emotion extremely effectively. This is technology used as an enabler. I've yet to come across someone using AI tech in the same liberating manner unlocking novel imaginary constructs in the way that those two did.
Take U2's October as a nice example. (You mentioned Zombie, incidentally one of my favorites, the anger and frustration in there never fail to hit me, I can't listen to it too often for that reason), superficially it is a very simple set of lyrics (8 lines I think) and an even simpler set of chords. And yet: it moves me. And I doubt any AI would have come up with it or even a close approximation if it wasn't part of the input. That's why I refuse to call AI generated stuff art. It's content, not art.