Kind of hard to interpret your FP, but that's how "the rush to FP" works, isn't it? The "it" in your Subject must mean "gibberish" or "commit the act of spewing gibberish", but I'm focusing on the "reason" part as an implicit question about why. On that basis I react negatively to your simplistic and absolutely dismissive answer. And I offer my Subject as an answer to your implicit question.
I do sort of agree with your negative sentiments towards LLMs, but I think they cannot be dismissed so easily for two main reasons. One is that so many people are eager to find an oracle and they eagerly want to believe whatever their favorite genAI says, up to the point of rationalizing hallucinatory gibberish. Never forget that "People believe what they want to believe." (I think the first time I heard that was from Dr Bill Martin about 50 years ago, but it's still true and even seems to be getting worse.)
The second reason is that the blending effect of the massive input tends to make the LLM output sound plausible just because of its similarity to "the center of mass of" the training data. Whatever the genAI says, it always sounds like something you've heard before. Or more likely like many things you've heard many times.
But my reaction was that the specific problem of this story is more like a brainworm. Usually that's a song or musical phrase that gets stuck in your head, but it can be an image or idea that becomes a kind of fixation. Therefore I am quite interested in how these brainworms are triggered because the triggers might work on humans, too. If so, I'm sure the applied psychologists are already hard at work figuring out how to abuse us with that new technique. Least malevolently for selling us a different brand of toothpaste but at the worst levels pushing stuff like building support for civil wars. Can you imagine "We've got to sink the Bismark" mutated to "We've got to sink Chicago"? (Did I just trigger a brainworm? Long time since I've heard that song...) [And then I searched this discussion for any mention of brainworms, but there was only a kind of reverse reference in a sig...]
By the way, I also think this topic is related to how facial recognition can be poisoned with a recognition icon. Some years ago they discovered how to make a small image that when included in a larger image would trigger recognition of a particular person without regard to the rest of the image. My memory is already fuzzy, but they essentially distilled the signature of a person and then included that signature in any input image to create a high probability match without regard to the actual faces that actually appeared in the image. And now I just thought of a funny game to play with the technique. Put one on your forehead, a different one on your right cheek and a third on your left cheek and you become three people depending on which angle they look at your face. Scatter them about and you could become a multitude?