The Em Dash is Gone: Sam Altman Confirms ChatGPT Finally Dumps Its Most Annoying Habit

OpenAI CEO Sam Altman has confirmed via his personal X account that ChatGPT has finally addressed one of its most frequently criticized “bad habits.” If users now specify in Custom Instructions that they do not want ChatGPT to use the em dash (—), the model will better respect that preference, reducing one of the more obvious telltale signs of AI-generated text.

Previously, no matter how clearly users phrased their prompts, ChatGPT seemed to ignore such requests and continued to rely heavily on em dashes in its responses. The overuse of this punctuation mark had already become, for many, a convenient heuristic for guessing whether a piece of writing had been produced by an AI.

As a result, whenever a text made excessive or pervasive use of em dashes, readers often began to suspect AI authorship, even though not all such writing was necessarily machine-generated. This distinctive stylistic quirk had effectively turned into an informal signal of possible AI involvement.

It remains unclear why generative AI models are so prone to sprinkling their output with em dashes, but one prevailing theory points to their training data. Large language models are trained on vast collections of books and online content—scientific papers, public forum posts, news articles, and more—where em dashes may appear with high frequency. Since the training process never explicitly flags this punctuation as something to limit, the model learns to reproduce it liberally.

With OpenAI’s latest adjustment, however, it will clearly become more difficult to judge whether a text is AI-generated based solely on an abundance of em dashes. Going forward, readers may need to scrutinize online content—and even printed material—with greater care when trying to determine how it was written.

Related Posts:

• Beyond GPT-5: Inside Sam Altman’s Multibillion-Dollar Plan to Build an AI Empire
• ChatGPT Reaches 800 Million Weekly Users, Cementing Dominance in Generative AI Adoption
• Sam Altman Declines Musk’s OpenAI Bid, Considers $9.74B Twitter Acquisition Instead
• OpenAI’s “X” Rival: Social Network Plans Emerge