OpenAI has updated ChatGPT to better comply with user requests to avoid em dashes, a punctuation mark that became widely associated with AI-generated writing. For months, users reported that ChatGPT would continue to insert em dashes even when explicitly asked not to, making the symbol a recognizable sign of chatbot-produced text across emails, school assignments, customer support chats, and online posts.
CEO Sam Altman announced the fix on X, calling it a “small-but-happy win.” OpenAI says the improvement works through custom instructions in ChatGPT’s personalization settings, allowing users to specify that the model should not use the punctuation. The company noted that the update does not remove em dashes by default but gives users more consistent control over their appearance.
OpenAI acknowledged the issue in a Threads post, where it jokingly apologized on ChatGPT’s behalf for “ruining the em dash.” The update reflects ongoing efforts to refine stylistic consistency in large language models as users increasingly seek more natural, personalized outputs.
For related coverage, read OpenAI Rolls Out Pilot Group Chat Feature for ChatGPT.
