No shit.
They talk about artificial "intelligence", "reasoning" models, "semantic" supplementation, all that babble, but it's all a way to distract you that large language models do not think. Their output does not show signs of reasoning, unless you're a disingenuous (or worse, dumb) fuck who cherry picks the "hallucinations" out of the equation.
And even this idiotic "hallucinations" analogy is a way to distract you from the fact that LLMs do not think. It's there to imply their reasoning is mostly correct, but suddenly it "brainfarts"; no, that is not what happens, the so-called hallucinations are the result of the exact same process as any other output.