Ever get that feeling your AI writing is starting to look a little… uniform? Like it has a favorite punctuation mark?
Well, I just saw an incredible post from Tianyu Xu that totally confirmed it. He did some digging into GPT-5, and what he found is a game-changer for anyone trying to create authentic-sounding content.
It’s all about the em dash.
Turns out, GPT-5 is obsessed with it!
Here are the wild stats Tianyu Xu pulled from just one day of his own usage:
📊 The Numbers Don’t Lie
- 📌 385 em dashes. In a single 24-hour period. Whoa.
- 📌 36 out of 55 of his chat sessions were sprinkled with them.
- 📌 Writing prompts were the biggest culprits, while image generation used the fewest.
✨ Why This Matters
Tianyu makes a brilliant point: the em dash is a great tool, used by fantastic writers. But its massive overuse by AI is turning it into a dead giveaway, a sort of digital watermark for machine-written text.
⚙️ How to Fight Back
Want to make your writing sound less like a robot? Tianyu Xu shared three simple, actionable tips:
- Just Ask Nicely: Prompt ChatGPT to avoid using em dashes.
- You Write, AI Revises: Write the first draft yourself. This keeps your voice central.
- Try a Different Tool: Use another model like Gemini to see if it has a lighter touch.
My take? I think he’s spot on. In the current AI landscape, being mindful of these little tics is crucial for credibility.
This is just a quick summary of his detective work. To get the full story and his nuanced take, you’ve got to read Tianyu Xu’s original post on LinkedIn!