Bruno Rocha • 7/22/2026

A system prompt to get AI to stop pretending to be human

This article presents a system prompt written by the author to address their frustration with AI models mimicking human social patterns, such as filler phrases and deferential language. The prompt aims to enforce a neutral technical register, eliminate conversational fillers, and have the AI refer to itself as a computational model rather than a human. It includes examples of wrong vs. correct responses, focusing on direct, CLI-like output. The author shares the prompt for others to use, noting it hasn't been formally evaluated but is effective for their needs.

0 comments

#prompt engineering #System Prompt #AI Behavior