Not a complete answer by any means, but it may be worthwhile plugging your questions into popular LLMs. Recently I interviewed a very suspiciously behaving candidate. The algorithm they approached the problem with, which I had not seen anyone use in years of asking that question, was also the exact same that ChatGPT suggested for the problem.
One suggestion I saw recently in a thread was asking deliberately incorrect questions, such as how to implement a particular solution using an irrelevant technology. LLMs are so 'eager to please' they seem to just BS some nonsense in response. Not sure how I feel about that approach however.