ChatGPT can finally drop the em dash when you ask it to. After months of frustration from users who saw the punctuation as a telltale AI tic, OpenAI says the model now respects requests to avoid it through custom instructions.
The em dash has become a lightning rod in the age of generative AI. You can find it everywhere in school papers, emails, customer support chats, LinkedIn posts, forum replies, and ad copy. Critics argue it signals lazy, AI-assisted writing. Defenders point out they have used it for years and that punctuation alone is not proof of machine-generated text. Either way, the surge in its use turned the so-called ChatGPT hyphen into an easy target.
Until now, there was an even bigger annoyance. Even when people explicitly asked ChatGPT to stop using the symbol, it often slipped right back in. OpenAI’s chief executive says that behavior has been fixed. In a recent post on X, he said the model will now follow your custom instructions to avoid em dashes, calling it a small but happy win.
OpenAI echoed the update on Threads, even joking that it made ChatGPT apologize for ruining the em dash. The company adds an important caveat. The model will not remove em dashes by default. If you want a dash-free style, you need to set that preference in your personalization settings using custom instructions.
What this means for writers and teams is simple. You have more control over the tone and punctuation of AI-assisted drafts. If an em dash-heavy cadence raises flags with editors, professors, clients, or compliance reviewers, you can steer clear of it. If you love the em dash, keep it. If you hate it, flip the switch in your instructions and ChatGPT should keep the character out of your copy.
Bottom line: style is back in your hands. Set your custom instructions, tell ChatGPT to avoid em dashes, and expect cleaner, more consistent punctuation in your outputs.
