OpenAI Says It’s Fixed ChatGPT’s Em Dash Problem

OpenAI says that ChatGPT will now stop using em dashes — if you ask it to.

The much-debated punctuation mark has become a recognizable “tell” in AI-generated text over the past year, appearing in school essays, LinkedIn posts, customer support chats, emails, ad copy, and even online comments. For many, the sudden ubiquity of the em dash became shorthand for “AI writing,” prompting criticism of overreliance on chatbots.

Others defended the punctuation, arguing that they’ve been using em dashes long before large language models made them popular. But because ChatGPT and other bots couldn’t seem to avoid using them — even when explicitly told not to — the so-called “ChatGPT hyphen” quickly became a red flag for AI-written content.

The issue has reportedly stumped OpenAI for months. Users complained that even when they added clear instructions, such as “avoid em dashes” or “do not use long dashes,” ChatGPT still inserted them into responses.

That’s now changing. In a post on X, OpenAI CEO Sam Altman announced,

“If you tell ChatGPT not to use em-dashes in your custom instructions, it finally does what it’s supposed to do.”
Altman called the fix a “small-but-happy win.”

Small-but-happy win:

If you tell ChatGPT not to use em-dashes in your custom instructions, it finally does what it's supposed to do! — Sam Altman (@sama) November 14, 2025

OpenAI followed up on Threads, jokingly forcing ChatGPT to “apologize for ruining the em dash.” The company explained that the model now honours style preferences set through custom instructions in personalization settings — meaning users who explicitly request no em dashes will see fewer of them in responses.

However, the update doesn’t remove the punctuation by default. Instead, it gives users more stylistic control over how ChatGPT writes — a subtle but symbolic fix to one of the model’s recognizable quirks.