Ever feel like your AI writing has a bit of a… dramatic pause? You know, the one I’m talking about.
Well, I just stumbled upon a super interesting post from Tianyu Xu, and he’s totally nailed GPT-5’s new favorite punctuation mark! I was blown away by the data he collected.
Tianyu ran the numbers on his own ChatGPT usage over 24 hours. The results are wild:
- 📊 He found a staggering 385 em dashes.
- 🤯 A whopping 36 out of his 55 chat sessions were sprinkled with them.
- ✍️ Unsurprisingly, writing prompts generated the most dashes.
Here’s the brilliant point Tianyu makes:
this overuse is turning the em dash into a dead giveaway for AI-generated content. It’s a shame, because it’s a great piece of punctuation, but he’s right that it’s becoming a red flag.
So, what’s the fix?
Tianyu Xu shares three practical ways to get your writing sounding more human again:
- 1️⃣ Just Ask: Simply tell ChatGPT, “Don’t use em dashes in your response.”
- 2️⃣ Write First, Revise Later: Draft the content yourself and then use the AI for polishing and editing.
- 3️⃣ Switch It Up: He notes that other models, like Gemini, don’t seem to have the same obsession.
Honestly, I think this is a fantastic observation. As Tianyu says, even though great writers use the em dash, we have to be careful not to let our work get flagged as robotic because of this AI quirk.
For the full breakdown and to share your own tips, go check out the original post from the man himself, Tianyu Xu!