By skyforbes 
In the past months, the em dash managed to become one of the most recognizable signatures of AI-generated writing. It would appear in school papers, emails, LinkedIn posts, customer service chats, ad copies and even internal business communication. Everywhere! As long as it was written by ChatGPT, you’d find em dashes there. And to make it worse, it was often fixed in places where human writers would not even use it.
The pattern became so common that the em dash itself became something like a detection tool, as people assumed that any paragraph containing one was probably written by ChatGPT. Now, many consider it a marker for ChatGPT content. I personally always found that annoying. I mean, this punctuation existed long long before LLMs and I found it disrespectful that its essence has somehow now been stripped to lazy writing.
No, em dashes were not invented by ChatGPT!!!
I admit those “It’s not your usual fashion show — it’s a runway revolution” kind of captions on Instagram truly scream ChatGPT. But actual writers use em dashes legitimately. To add rhythm. To add surprise. To add drama. To add personality. And I loved it. Oh, I loved that punctuation for the power it held. The flexibility. The perfectly timed pause. Imagine having to abandon all that just because of the bastardization. I mean, how do you explain to your readers and clients that the idea that the use of em dashes is a sign that a piece is just one giant slob of AI-generated text is false? That it is a terrible, mistaken idea and not the dead giveaway some people think it is?
What do you mean, “ChatGPT hyphen?” Such disrespect!
I always found it funny how those who came up with the idea (and the ones who swallowed it it hook, line and sinker) never stopped to think that if chatbots use em dashes, it’s because human writers use em dashes. It’s simple logic. I mean, LLMs are trained on large amounts of material and to start with, this material was the work of human writers.
Now, em dashes are everywhere in the training materials, from books to essays. It only makes sense that the bots, programmed to copy the data they are trained with, picked up on em dashes as a default natural flow. That explains why LLMs like ChatGPT are stubborn about the use of em dashes. No amount of prompt, setting or even warning could stop those punctuation cockroaches from creeping up.
So, the reputation that em dashes are a robotic thing stuck. As one of the many writers who stopped using em dashes to avoid our work being dismissed as AI’s at the first sight of the punctuation mark, I find it very ironic. After all, em dashes are one of the best tools writers have for not sounding robotic in the first place.
ChatGPT itself knows that it “flew too close to the punctuation’s sun,” as it admitted in an OpenAI post announcing that the issue has been fixed.
In another post on X, the company’s CEO Sam Altman said that ChatGPT would finally listen to a user’s request not to use the symbol when that preference is set inside Custom Instructions. A small but happy win, he called it.
The company echoed the announcement on Threads, explaining that users who add a preference in personalization settings will now see the model reliably avoid it (without bringing it back two sentences later).
While this fix is welcome (it’s never too late to go back to my beloved em dashes), it took OpenAI longer than expected. And I am still a little salty about it despite the apology letter. The poor punctuation took too much of a beating!
OpenAI has not provided any technical details about how the fix works. But it did clarify that ChatGPT’s writing by default will still contain hordes of em dashes. The difference now is that users have control over whether they want it in the content the AI generates for them or not.
The update is part of OpenAI’s growing push toward customization, as the company’s chatbot already offers memory, tone presets and experiments that allow users to fine-tune style, conciseness or warmth. Being able to control something as granular as a punctuation habit fits neatly into that strategy. It also addresses a larger trust issue. And I hope that more people take advantage of the update so that the em dash can stop being a red flag.