So you hate em-dashes now?

AI-generated content is getting harder and harder to distinguish a lot of the time. A bad side-effect of this is a distrust of what we see now—assuming we haven't given up on the idea of trustworthy content entirely. There are only a few things we can do to strip out AI slop, such as only reading or watching content published prior to 2021 (I've been watching Buffy The Vampire Slayer almost every night for two years, now, it's definitely a solid method). But if we want something up-to-date, like on social media, we can't filter it out, so all we can do is try to detect it and then skip past it. The problem is that it's much harder to discern between machine-generated garbage and human-generated garbage in short-form styles of content.

Naturally, there have been trends of advice that circulate as people come up with quick ways to detect the signature of AI. It used to be hands in AI-generated images: just look at the hands! If you're deeply disturbed, it's AI. Another one that still holds in video format is looking for discontinuity of environment or really short shots (1-10 seconds)—it's very difficult to generate long clips with AI, especially with a consistent, logical background setting. However, the clip length alone doesn't mean it's AI-generated, it's just a common data point we can use (foreshadowing!).

And what about for text/written format?

One of the ways people have (incorrectly) distilled down a method of detecting AI-generated content is by accusing the presence of em-dashes as being the sure sign of AI work. This is deeply saddening to me because as a typophile I adore the em-dash—to the point that I, myself, overuse them as you've likely noticed, ha! However, I have a theory for why AI loves to employ these long dashes and why non-designers are now discussing this character as though it's old hat.

Always Blame the Marketers...

Fresh out of college, I was determined to make all of my marketing material typographically impeccable. I had landed a side-gig where I would typeset a PDF version of a case study. All that means is that I would get one or two pages in a Google Doc and I'd put it into a template. I would meticulously adjust the kerning (spacing between letters) to avoid orphans (a word that gets left on its own line at the end of paragraph), painstakingly change out-of-the-box prime marks ('/") into real quote marks (‘/”), and—with greatest love and care—change double-hyphens into em-dashes (—).

No one cared that I did that.

But I sure did. If nothing else, it was my way of doubling-down on quality and practicing all of the design skills I had learned in college. I would continue to do these things even as I morphed into a web developer. I've had to format many, many thousands of pages of content, and I still want that same high standard of typographic quality.

With all of this exposure to marketing content, and with my hyper-fixation on characters like the em-dash, my personal theory on why AI uses them so much is simply because of marketing. AI was trained largely from the Internet because it's such a huge source of already-digital, publicly-accessible content (ignoring copyrights, of course). And what kind of content is on the Internet? MARKETING. It's everywhere. It's probably the main kind of content that is out there—again, I know this largely because of having worked in marketing for so long, so there is some bias, but I think it's accurate to say at least half of all websites are marketing sites.

Inbound marketing, in particular, is a way that we've pumped content into the Internet that was ready to be devoured by AI. "Inbound" refers to a marketing strategy that attempts to attract search traffic to your website through content that is not explicitly marketing material. That's why companies have blogs (still), and sometimes they've got more in-depth content throughout the site. Even AI companies have content sections that aren't technically marketing, like research or guides, and these may even be very useful. As long as you're getting to the site, it's ok if you're not reading a product page, you're getting a taste for the brand, becoming familiar with the site and the names of the company/products, maybe even starting to trust the company because of the content they've provided to you.

All that to say, I think a huge amount of the content that has ended up training AI models comes from marketing websites, and marketers have a little bit of a style that then comes through what AI produces. For better or worse, marketers love em-dashes—especially em-dashes that appear in headlines and bulleted lists:

• Everything ready—before you are!
• Become the smartest in the room—without breaking a sweat.

This short, interruptive sentence structure is everywhere on marketing sites, even if it's just the bottom call to action after a well-written article that isn't so short and punchy. Over the years, my marketing content detection device (my brain) lights up like a Christmas tree on almost every response I see from LLMs.

Em-dashes Are, Like, Popular Now or Whatever

So, yes, em-dashes have been plaguing the Internet for many, many more years prior to the ability of an AI model to string coherent words together. Marketing has a style that has passed to LLMs, probably because of its overrepresentation in the training data. Why in the world are non-designers talking about em-dashes so much?

I think that the discussions around AI-generated content have brought the em-dash into public consciousness. If you're not a student, and especially if you're not into writing, you probably have never even noticed the em-dash. Maybe you have barely noticed just from reading books, but it's passed over like any other punctuation mark like commas and periods. I can count on two fingers the times in my life I've heard someone talk about diacritical marks in casual conversation, and never have I heard talk of virgules or manicules or any number of other glyphs.

In my estimation, it just so happens that as people have become aware of the need to detect AI-generated content, we've finally noticed all of those em-dashes floating around and realized that AI overuses it. Now the name of this character is infamous and I've heard some people say they'll never use them because they've incorrectly determined em-dashes to be an AI-only style choice. This is terribly sad to me and I refuse to stop using em-dashes if for no other reason than my reverse-hipster impulse (I was using them before it was uncool).

Instead of avoiding em-dashes like the plague, I prefer instead to formally introduce you to this beautiful mark, which has been used for hundreds of years!

If you want to impress your designer friends, you can use these keystrokes to type an em-dash:

Even more impressive is knowing why we call it the em-dash, and for that, we need to look at printing presses, which used to be typeset by hand. Letters used to be cast in metal blocks that would be placed together to build up words, sentences, and eventually paragraphs and pages. As James Craig and Irene Korol Scala explain in Designing with Type, 5th Edition, we see this "em" naming come through:

Em-quad: In handset type, a metal space that is the square of the type body size; that is, a 10-point em-quad is 10 points wide. The em gets its name from the fact that in early fonts the M was usually cast on a square body. —page 164

The capital letter "M" is also one of the largest letters, so it makes for a good measurement of space relative to the typeface design.

Em-dash: Also called a long dash. A dash the width of an em-quad. —pg 164

And finally, we get a glimpse for when to use it:

The em-dash is the mark commonly meant by the term "dash." It indicates an abrupt break in thought or speech, and it may be used instead of commas or parentheses to set off a parenthetical clause. —page 157

Don't believe just a design book? Here are two other resources that really help. The first is A Pocket Style Manual, fourth edition, by Diana Hacker, which is still my go-to when I need help remembering specific grammar or style rules (even though my writing is still all over the place, haha). However, Hacker does explain that you can use the double hyphen (--), but since it's 2026 and em-dashes even come on Android phone keyboards now, there's no excuse for using that style which is left over from typewriter days (see next resource as well).

The dash may be used to set off material that deserves special emphasis. When typing, use two hyphens to form a dash (--), with no spaces before or after them. (if your word processing program has what is known as an "em-dash," you may use it instead, with no space before or after it.)

Use a dash to introduce a list, a restatement, an amplification, or a dramatic shift in tone or thought. ... Use a pair of dashes to set off parenthetical material that deserves special emphasis or to set off an appositive that contains commas. ... CAUTION: Unless you have a specific reason for using the dash, avoid it. Unnecessary dashes create a choppy effect. —page 84-85

Finally, we have my favorite typography resource, The Mac is not a typewriter, second edition, by Robin Williams. Her whole premise is using the advanced tools of the computer to overcome bad habits learned by typewriter limitations. One of those limitations was the lack of any other dash (en or em), so typists were forced to use the double-hyphen instead of the em-dash.

The em dash is twice as long as the en dash—it's about the size of a capital letter M in whatever size and typeface you're using at the moment. This dash is often used in place of a colon or parentheses, or it might indicate an abrupt change in thought, or it's used in a spot where a period is too strong and a comma is too weak... ... Since you were properly taught, of course, you know that the double hyphen is not supposed to have a space on either side of it—neither is the em dash, as you can see right here in this sentence. page 24

I hope this was as enlightening as it was nerdy, so that you can now be annoyed when other people dismiss the em-dash as something that only AI uses, lol. But really, I just want to see more good typography out there, and the em-dash is a natural—and beautiful—way to represent grammar. AI over-uses emojis, em-dashes, and other marketing-style sentence structures, so while those are good factors to use in detecting AI content, it doesn't mean it is. At least here at CYBORG_, you know it's human-written despite my overuse of my favorite character.