GenAI is a threat, because language is compression.
It's valid to treat the exact same statement made by a middle-schooler and by a seasoned professional differently. Because one cannot possibly have the same substantive intent as the other - i.e. they decode to different things. Obviously depends on the statement, nothing over "I like chocolate".
Likewise with ChatGPT-etc. ChatGPT can reason, and has knowledge, but both are very poor. What matters is reasoning - it's not just lacking, it's "noisy", hence self-inconsistent. Humans have much a greater source of inconsistency, but in a fundamentally different way - emotion - that's better or worse than GenAI depending on subject.
Nature of poorness aside - the extent ("a great ill", "a bit dishonest") of descriptors is the most pertinent, to my mind, source of difference due in interpretation when reading GenAI content. Every adjective is an evaluation, and human evaluation is a combination of pure reason, emotional reason, and memory - memory formed by experience to great extent.
It is, so far, a small or non-issue to subject experts. But what if one's unfamiliar? We're forced to take the speaker's word for some things, and this is far less credible for GenAI. In the future, issues will grow more nuanced. Yet, in enough-future, they're bound to disappear and GenAI reasoning will be superior to humans' in nearly absolute terms. The shortcomings might be, for yet-unknown reasons, the supremacy of analog computation - e.g. human-like consciousness may be impossible digitally. Brain-computer interfacing or genetic engineering can also turn tides.
Of course GenAI is also very helpful, as other answers describe and-more.