OpenAI, an American technology company, recently announced that users can now stop ChatGPT from automatically using em dashes in its output through personalized settings. In recent years, controversy has arisen after people discovered that dashes appearing frequently in school papers, emails, comments, customer service chats, LinkedIn posts, online forums, and advertising copy have become a "signature feature" of artificially generated text. Some critics believe this is a sign of lazy writers relying on AI.
However, many people defend the dash, pointing out that it was a common punctuation mark in many writing styles long before generative large models became commonplace. However, because chatbots inevitably use dashes frequently, "ChatGPT dashes" have become a new hook for AI content in the eyes of many readers, although this cannot be used as absolute evidence of AI-generated content.
This problem once made OpenAI quite troublesome. Even if the user explicitly asks ChatGPT not to use dashes in the prompt, the previous system still cannot completely avoid the symbol.
Now, OpenAI CEO Sam Altman posted on the X platform that the company has fixed this problem. "If you ask ChatGPT not to use dashes in your custom directive, now it finally does so," he said. Altman called this a "small but pleasant improvement."
OpenAI also added on the Threads platform that ChatGPT will better comply with instructions not to use dashes after users enter relevant requirements through personalized custom instructions. However, this adjustment will only take effect after explicit instructions from the user, and dashes will not be automatically disabled by default. But users will have more control over how often dashes appear in generated content going forward.