LLMs write by picking which next word best correlates to a response from the prompt, so they tend to follow grammar extremely well but make logical mistakes.
Humans write by forming their thoughts into word sounds, then transcribing them, but a single pronunciation can have multiple spellings, depending on the context, but the context is somewhat abstracted by that point, so humans regularly write homophones or malapropisms of the appropriate word.
LLMs write by picking which next word best correlates to a response from the prompt, so they tend to follow grammar extremely well but make logical mistakes.
Humans write by forming their thoughts into word sounds, then transcribing them, but a single pronunciation can have multiple spellings, depending on the context, but the context is somewhat abstracted by that point, so humans regularly write homophones or malapropisms of the appropriate word.