Ever feel like your ChatGPT responses are looking… a little samey? 🤔
I’m not talking about repeating words, but a specific piece of punctuation. This little guy: the em dash.
Well, I just saw this incredible post from Tianyu Xu, who actually put numbers to this feeling. The results are wild!
He ran the stats on his own usage and discovered GPT-5 gave him 385 em dashes in a single day. That’s insane! He found that 36 out of his 55 chat sessions were littered with them.
✨ The Big Problem
Tianyu makes a brilliant point. The em dash is beautiful, but its overuse is becoming a major tell for AI-generated content. As he puts it:
this beautiful punctuation mark is now an “indicator of AI-written content.”
Even if you’re a human who loves using them, your writing might get unfairly flagged.
⚙️ How to Avoid the Em Dash Takeover
Luckily, Tianyu Xu didn’t just spot the problem, he offered clear solutions. Here are his top three methods to get your writing back to normal:
- Just Ask: Simply tell ChatGPT, “Don’t use em dashes in your response.”
- Revise, Don’t Write: Write the first draft yourself and use AI for polishing and revision, not for initial generation.
- Switch Models: Try using another AI, like Gemini, which might not have the same stylistic tic.
I think he’s totally right. In this new AI world, being aware of these little tells is a massive advantage.
This is a super sharp observation! For the full breakdown and his original thoughts, you have to go read Tianyu Xu’s full post on LinkedIn. He’s really onto something here.