Particle.news
Download on the App Store

OpenAI Lets ChatGPT Drop Em Dashes on User Request

The change activates through custom instructions, signaling a shift toward finer-grained personalization.

Overview

  • CEO Sam Altman said on November 14 that ChatGPT will honor requests to avoid em dashes when saved in custom instructions, calling it a "small-but-happy win."
  • OpenAI clarified on Threads that this is a personalization setting rather than a default change to the model’s writing style.
  • The em dash had become a widely recognized tell for ChatGPT-generated text, frustrating users who felt the punctuation made their writing look automated.
  • Some users reported the fix had not reached their accounts or that the chatbot still occasionally inserted the mark.
  • The update aligns with OpenAI’s broader effort to let users steer tone and formatting through features like memory and custom GPTs.