AI Can Pry the EM Dash Out of My Cold, Dead Hands

The em dash. Odds are you’ve heard of it — and maybe you’re offended by the way I just used it. You wouldn’t be alone. 

In 2026, the em dash has become a symbol of AI slop. A red flag that something was written by a machine and not a human. Suddenly, everyone has an opinion on this punctuation mark. Most of those opinions are: stop using it.

And I’m here to say: over my dead body.

I’m a professional writer and I refuse to give up the em dash — and you should too. Here’s why.

The Great Em Dash Scare of 2025

A few years ago, no one knew what the em dash was. Not unless you were an editor or proofreader, anyway. I know, because I was one. 

Before my ad career, I edited books for a living. I spent many, many hours replacing misused hyphens, en dashes, colons, and semicolons with the proper punctuation point: the em dash.
Everyday people didn’t know what an em dash was. Even serious writers had a tendency to misuse it. And when I took my first copywriting role in 2016, clients certainly didn’t care what kind of dash I was using. 

Then along came ChatGPT, and the rumor that, allegedly, the LLM “overused” the em dash. IMO, this is debatable. I’ve been tinkering with ChatGPT since the early days (not my preference, but as a writer and marketer, it’s my job). I have many, many criticisms of Chat’s work, but not once did I think it was too dash-happy. 

But somehow, someway**, it became Internet Fact that the best way to tell AI writing apart from human writing was the em dash. 

Fast-forward to today. I’m fielding questions and concerns about the em dash every week. Should we use it less often? Stop using it altogether? Can we use a hyphen instead? (Absolutely not, btw.) How about a semicolon? (Sure, if you want to sound like a 19th-century Transcendentalist.)
 
These are all fair questions as we collectively grapple with a struggle that is becoming increasingly more difficult: telling apart what’s fake from what’s real. 

But we, as a collective, seem to have skipped the most important question here.

Is the Em Dash Even a Good Way to Prove What’s AI-Written?

The short answer: no. 

The medium answer: if the only thing differentiating your writing from AI slop is a dash, punctuation is the least of your problems.

And now for the long answer.

Why Giving Up The Em Dash Won’t Solve Your Problems

First of all, AI is run by corporations with a lot to prove. Mostly that their AI is as smart as a person, sounds like a person, and can replace people. If you think these companies haven’t already heard the complaints about em dashes and made tweaks to their algorithms, you’re wrong. 

If Chat ever did overuse the em dash, it surely isn’t anymore. So you might as well keep using it. It’s a perfectly good punctuation point. Ironically, it makes writing sound a lot more human — a casual stop mid-sentence to add emphasis or change topics emulates the way we talk every day.

And if Chat did overuse the em dash? Okay, fine. It’s going to move on to something else next.
 
Trying to keep up with the next trend in AI slop will be a losing game. There will always be a new “tell” that AI content isn’t real. (Until, of course, we reach the Singularity. I’d love to think by then we’d have some regulation of AI content, but I have little faith in humanity these days, so I’m guessing by then we’ll have bigger problems than telling whether ad copy was written by AI, like, idk, robot overlords.)
 
Secondly, and most importantly, don’t forget: LLMs are only advanced predictors of the next most likely sequence of words. As predictors, they’re, well, predictable. AI copy is generic and unsurprising by nature. Not only does AI scrape existing writing from the internet to make its decisions, but it’s also limited by parameters that tell it not to create content that could be offensive, harmful, off-brand, or sensitive. 

Not that your writing should be any of these. But the best writing takes risks. The best writers know how to bend the rules without breaking them. A prime example: you can’t make a joke without it being at someone's or something’s expense. A human can decide how far to take it, and when it’s appropriate. AI can’t — a good reason why it’s not really allowed to. If you’ve ever wondered why AI’s sense of humor is so bad, there’s your answer. 

True creativity is a skill that takes nuance, empathy, calculated risk. Oh, and failure — something corporations would never want to do in front of their investors. For all of these reasons, creative thinking is something only a human can do. At least for now.

The only surefire thing you can do to differentiate your human-made writing from AI is by writing creative, painstaking, unexpected, risky stuff — which you should be doing anyway. 

*While this is an easy way to remember the name of the em dash and tell it apart from its shorter counterpart, the en dash, this isn’t how these punctuation marks got their names. They’re a holdover from the print era, where em and en were units of measurement on metal type pieces. 

**How that rumor became widely accepted as truth isn’t clear, but I have my theories. One is that the people who used ChatGPT weren’t the kind of people who typically used em dashes. That is to say, they weren’t very good writers in the first place. Of course they saw this strange punctuation point and said, “Huh???? No one uses this thing!” Of course they knew that the sudden appearance of advanced punctuation in their work would look sus.