OpenAI Says It Has Fixed ChatGPT's Dash Issue

OpenAI has announced that ChatGPT will now comply if you explicitly instruct it to avoid using em dashes—a punctuation mark that had become emblematic of AI-generated text. Over recent months, these dashes have appeared pervasively across school essays, emails, customer service chats, LinkedIn posts, online forums, ad copy, and more, drawing criticism for signaling lazy writing or overreliance on AI tools.

While many users have defended their use of em dashes—arguing they were part of their personal writing style long before large language models (LLMs) adopted them—the persistent inclusion by chatbots made the so-called “ChatGPT dash” an increasingly unwelcome hallmark in digital content, even though it’s not a definitive indicator of AI authorship.

This issue had frustrated OpenAI for some time, as users previously couldn’t get ChatGPT to stop using em dashes—even when directly asked to do so.

Now, according to CEO Sam Altman, the problem has been resolved. In a post on X (formerly Twitter), Altman wrote: “If you tell ChatGPT in your custom instructions not to use em dashes, it will finally listen.” He described the fix as a “small but satisfying win.”

In a separate post on Threads—where the company even had ChatGPT issue an apology for “breaking the dash”—OpenAI clarified that users can now better control dash usage by specifying this preference in their Custom Instructions under personalization settings. This update doesn’t eliminate em dashes from outputs by default, but it does give users greater control over how often they appear.