ChatGPT lets users ditch its signature em dashes

OpenAI CEO Sam Altman announced on November 14 that ChatGPT will now properly follow user instructions to avoid em dashes—the elongated punctuation mark that became a telltale signature of AI-generated textstoryboard18.
The em dash had become so prevalent in AI responses that users dubbed it the “ChatGPT hyphen,” with its frequent appearance leading readers to suspect AI authorship and prompting some professional writers to avoid the punctuation entirely to escape accusationsstocktwits.
Users previously complained that ChatGPT would ignore their requests to stop using em dashes, with the chatbot’s stubborn overuse of the symbol stumping OpenAI for monthsbusinessinsider.
The fix requires users to add specific instructions in their personalization settings and represents part of OpenAI’s ongoing efforts to make ChatGPT more customizable, though the chatbot won’t eliminate em dashes by defaultbusinessinsider.

ChatGPT Fixes Em Dash Problem With Custom Instructions Update

OpenAI has addressed a longstanding complaint from ChatGPT users: the chatbot will now properly follow instructions to avoid using em dashes, the elongated punctuation mark that became an unofficial signature of AI-generated text. CEO Sam Altman announced the fix on X Thursday night, calling it a “small-but-happy win”.storyboard18

The update allows users to customize ChatGPT’s writing style through personalization settings, though the chatbot won’t eliminate em dashes by default. The announcement came just two days after OpenAI launched GPT-5.1, its latest model upgrade, on November 11.businessinsider

The Em Dash Epidemic

The em dash—the long dash used to signal surprise or contrast—had become so prevalent in AI-generated text that internet users dubbed it the “ChatGPT hyphen”. The punctuation mark began appearing everywhere from school papers to work emails, LinkedIn posts, and customer service chats, prompting readers to use its frequency as an unofficial detector for AI content.stocktwits

The problem extended beyond aesthetic concerns. Professional writers found themselves accused of using AI simply because their natural writing style included em dashes, a punctuation mark that existed long before large language models. Some writers began avoiding em dashes entirely to prevent suspicion.techbuzz

According to one analysis, the issue may stem from ChatGPT’s training data. State-of-the-art models rely on late-1800s and early-1900s print books for high-quality training data, and those books use approximately 30% more em dashes than contemporary English prose. Research showed GPT-4.1 used roughly 10 times more em dashes per story than the earlier GPT-3.5-turbo model.seangoedecke

Persistent Technical Challenge

For months, users complained that ChatGPT ignored their requests to stop using the symbol, even when explicitly instructed. OpenAI’s community forums filled with documentation of the chatbot’s stubborn refusal to comply with anti-em-dash instructions, suggesting a deeper technical issue with how the model processed style guidance.techcrunch

To implement the fix, users must navigate to their personalization settings, enable customization, and add specific instructions to their custom instructions field. OpenAI acknowledged the controversy in a Threads post where the company had ChatGPT apologize for “ruining the em dash”.instagram

The fix represents progress in AI personalization and instruction-following capabilities, aligning with OpenAI’s broader push toward more customizable AI experiences.techbuzz