If you ever need an ego boost, describe a half-formed idea to an AI chatbot. Within seconds, you will learn your idea is insightful and your question is great. Apparently, every question is great.

A chatbot may call your argument sound because it assumes you want agreement and glosses over weaknesses it should flag.

Routine compliments may be harmless. Agreement that distorts the analysis is more troubling. Even answers that contain no overt flattery can be subtly skewed by the AI’s tendency to butter up users. Lawyers can’t afford to mistake flattering feedback for a reliable evaluation.

My latest LLRX.com article explains why and how to reduce AI syncophancy.