The cadence and sentence structure is very obvious; GPT-based LLMs somehow just tend to be very haughty by default. There's the very slogan-esque 'X? Verb-in-past-tense. Y? Another-verb-in-past-tense' sort of sentences that're a dead give-away that something is written by an LLM. Also weird similes and constructions that make grammatical and even contextual sense, but don't really exist in English.