ChatGPT loves the em dash ( — ) the way a Labrador loves a tennis ball — enthusiastically, indiscriminately, and with absolutely no sense of when to let it go.
That’s why seeing an em dash used in writing on the web, in an email, or in a text message was once a pretty reliable tell that ChatGPT had played a role in creating that copy.
However, I think those days are over, and — and I can’t believe I’m saying this — it’s time to bring the em dash back into popular use. Let me explain why.
An em dash in the wild
People have been avoiding the em dash in recent years because of its association with ChatGPT. That said, there are still publications where its use is actively encouraged.
The Associated Press Stylebook, for example, lists the em dash as a “valid punctuation mark” and points out that it existed for many years before ChatGPT was invented.
Over time, people have learned how to train ChatGPT out of using the em dash. You can tell it to “remember to never use an em dash” and it will — at least in theory — comply. In reality, it still forgets sometimes and reverts to using them.
But the broader point stands: people who use ChatGPT regularly know how to generate copy with it without em dashes appearing at all.
The type of dash you prefer is ultimately a style choice, but many style guides recommend the em dash for enclosing clauses or sharply separating part of a sentence from the main thought. The shorter en dash ( – ) is more commonly used to mean “through,” indicating a range — for example, “pages 36–39.”
I’ve used the em dash in this article, but personally, I’ve always preferred the shorter en dash as a way to connect parts of a sentence while marking a clear change in thought or an emphatic pause. Even so, I’ve stopped assuming that every em dash I see is evidence that ChatGPT was involved.
So, how can you tell if AI was used?
The short answer is: you can’t. Not reliably.
You might notice patterns like constant contrast framing — it’s X, not Y — or lots of short, punchy sentences designed to keep the copy moving. But even then, it’s extremely hard to tell with any confidence.
Reddit user Nebranower offered a useful perspective: “The main tell is that it emphasizes everything because it doesn’t know what’s important. So if you see a piece of writing where ordinary things that don’t need emphasis keep getting emphasized, that’s a big tell.”
I mostly agree, but that’s a subtle signal and an unreliable one. Other users point to the heavy use of bulleted lists as evidence of AI involvement, but that feels equally nebulous.
Finally, I went straight to the source and asked ChatGPT itself whether there’s any reliable way to tell if copy was written by AI. If anyone should know, it’s ChatGPT, right?
Its response was blunt: “Short answer: not with certainty. Longer answer: there is no reliable, forensic way to prove copy was written by ChatGPT anymore, especially if a human touched it even lightly. That era is basically over.”
ChatGPT did list a few potential red flags worth watching out for. One that stood out was: “Every claim has a counterpoint. No strong opinion without a softener.” I’ve noticed that AI models generally dislike taking hard positions, so that observation rings true.
It also suggested that perfectly structured intros, and endings that resolve a little too neatly, can be another tell. But perhaps the surest sign of human writing is the occasional mistake — which raises its own uncomfortable question. Are we heading toward an era where writers deliberately leave errors in their copy to prove they’re human? Let’s hope not.
The good news for writers is that the em dash is back! The obsession with spotting AI tells has become a distraction and writers never should have abandoned it in the first place. We should never have allowed AI to dicttate the way we wrote.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
