Whoa. Ever feel like your AI-generated text has a weirdly specific quirk you just can’t put your finger on?
I just saw this incredible post from Tianyu Xu that totally nailed it. He’s been tracking his ChatGPT usage and discovered something fascinating: it’s absolutely obsessed with the em dash (—).
Check out the stats Tianyu Xu pulled from just 24 hours of use:
- 🤯 He found a staggering 385 em dashes.
- 📊 36 out of his 55 chats contained at least one.
- ✍️ He noticed that while image prompts had the fewest, writing prompts were littered with them.
Here’s why I think this is a big deal:
The em dash is quickly becoming a dead giveaway for AI-written content. As Tianyu points out, it’s a shame because it’s a great punctuation mark, but its overuse now makes writing feel robotic.
So, how do we fix it?
Tianyu Xu shared a few brilliant (and simple) methods to get your writing sounding human again:
- 📌 Just Ask: Add a simple instruction to your prompt like, “Do not use em dashes in your response.”
- 📌 Write First, AI Second: Draft the content yourself and then use ChatGPT for revisions. You stay in the driver’s seat.
- 📌 Try Another Model: If you’re still stuck, he suggests trying a different AI like Gemini, which has its own unique writing style.
This is a super sharp observation. It’s a great reminder to be mindful of how we use these tools to keep our voice authentic.
This is just my quick breakdown, so go read Tianyu Xu’s original post for the full discovery!