OpenAI has announced that it has resolved a long-standing issue with ChatGPT’s overuse of em dashes, a punctuation mark that has become a telltale sign of AI-generated text. The em dash—often used in place of commas, parentheses, or colons—has appeared frequently in content produced by the chatbot, leading to widespread criticism. Many users and educators have pointed to its repetitive use as a red flag that a piece of writing was likely generated by an AI, especially in school assignments, emails, social media posts, customer service messages, and even professional content like LinkedIn updates and ad copy. While some defenders of the em dash argue that it’s a legitimate and stylistically valid punctuation tool with a long history in writing, its overuse by AI models has made it a symbol of machine authorship. This has led to growing frustration, with critics accusing users of relying too heavily on AI tools and producing content that feels artificial or lazy. For months, users were unable to get ChatGPT to stop using em dashes, even when explicitly instructed to do so. The issue stumped OpenAI’s team, as the model continued to default to the punctuation despite user preferences. Now, OpenAI CEO Sam Altman has confirmed the fix. In a post on X, Altman called the update a “small-but-happy win,” noting that “if you tell ChatGPT not to use em-dashes in your custom instructions, it finally does what it’s supposed to do.” The change is part of a broader effort to give users more control over the tone, style, and formatting of AI-generated content. In a follow-up post on Threads, OpenAI explained that the model will now respect user preferences when it comes to em dash usage—specifically, if the user disables the feature in the custom instructions within their personalization settings. This means the em dash won’t be eliminated across the board by default, but users can now choose to avoid it entirely when needed. The update marks a step toward making AI-generated text more customizable and less predictable, helping users produce content that better aligns with their intended style. While the em dash may still appear in some outputs, the new control gives writers the power to fine-tune the model’s behavior and reduce the risk of their work being flagged as AI-generated.