Given the extensive array of examples from GPT-3's experiments, I'd say it's perfectly possible that those dialogues are real. Yet, I agree it's important to always doubt about this kind of news - and having the criteria to doubt is equally as important.