ChatGPT will now stop using em dashes if users ask, says Sam Altman

OpenAI has finally addressed one of the quirks that has frustrated ChatGPT users for years: the chatbot’s habit of sprinkling em dashes throughout its responses.

For months, people noticed that the punctuation mark, used to connect ideas or add emphasis, was appearing in almost every ChatGPT reply. It became so common that some began calling it the “ChatGPT hyphen.”

Its frequent use sparked criticism, with readers assuming anyone relying heavily on the em dash was lazily leaning on AI to write their content. This trend was visible everywhere, from school assignments to LinkedIn posts, emails, online forums, and customer service messages.

While many writers defended the em dash as a legitimate stylistic choice, the fact that ChatGPT could not avoid using it made it feel like a telltale sign of AI-generated text. Users repeatedly asked the chatbot to stop, but it seemed unable to follow those instructions, until now.

OpenAI CEO Sam Altman confirmed the change on X, writing that if users set a preference in ChatGPT’s custom instructions asking the AI to avoid em dashes, it will now respect that choice. He called the update a “small-but-happy win.”

According to OpenAI, the update gives users more control over the punctuation ChatGPT uses. The AI won’t automatically remove em dashes from every response, but if instructed through the personalization settings, it will avoid overusing them. The company also shared a note on Threads, where it acknowledged the em dash issue and even apologised for “ruining the em dash” in ChatGPT’s earlier responses.

This update may seem minor, but for users who have long been frustrated by the unrelenting em dash, it represents a small victory and a step toward giving people more control over how AI writes for them.