Overview
- Sam Altman said on X that ChatGPT now honors requests to avoid em dashes when the rule is set in custom instructions.
- OpenAI showcased the change on Threads with a playful apology, clarifying that it is a user-configurable preference rather than a default removal.
- Early feedback shows an uneven rollout, with some users reporting that their chats still produce occasional em dashes despite the setting.
- The update aligns with OpenAI’s emphasis on personalization features such as memory and custom GPTs, as well as improved instruction-following in recent model updates.
- The em dash had become a widely cited tell for AI-written text, drawing complaints from writers and editors even as others defended its traditional use.