We often search for subtle clues to distinguish human-written text from AI-generated content. Yet, recently, one punctuation mark unexpectedly stole the spotlight: the em dash. ChatGPT seemed to use it with surprising frequency—even when explicitly instructed otherwise.
This peculiarity sparked a strange consensus across the internet: the belief that humans simply don’t use dashes. Tech forum users labeled them a ‘GPT-ism,’ a tell-tale sign of robotic writing that supposedly didn’t align with contemporary communication styles. One user on an OpenAI forum even voiced concern that the abundant dashes hindered ChatGPT’s use in customer service, making AI detection too easy. It was puzzling how many confidently asserted that no real human would ever employ this punctuation, and that anyone who did would risk being mistaken for a machine.
Naturally, this infuriated those of us who do use dashes. As a former proofreader, I could passionately elaborate on the nuances of even the lesser-known en dash. While I might concede my fondness for dashes is atypical, I was genuinely shocked to hear their very existence as a human tool questioned. The dash is a venerable and perfectly normal part of sentence construction! Literary giants like Dickens, Dickinson, and Nietzsche, along with Stephen King novels and even this very magazine, are filled with them. In fact, one reason for their popularity is how they can evoke a more casual, natural, speech-like rhythm compared to colons, semicolons, or parentheses. We don’t articulate in perfectly formed sentences; we think and speak in interconnected thoughts, which frequently interrupt, introduce, and intertwine, forming richer, more intricate ideas. (Sometimes, they don’t quite connect, as J.D. Salinger’s dash-laden dialogue beautifully illustrates how often our thoughts remain unexpressed or incomplete.) This is precisely the function of punctuation.
However, the true AI indicator isn’t just the dash itself, but its orthography. ChatGPT renders its em dashes in the classical printed book style—a stroke the width of the letter ‘M’ with no spaces around it. This isn’t how most computer users type. Many might not even know the specific keystrokes for this character (some mistakenly called it a ‘ChatGPT hyphen’). Instead, the typical user types a single hyphen (-) or a double hyphen (–), often corrected by software to the more subtle en dash (–). Crucially, most online users and publications add spaces around their dashes, a practice that aids text wrapping and readability on screen.
Despite these technical distinctions, the debate persisted, fixating solely on the dash itself. It was discussed as if it were some strange, archaic symbol no self-respecting human would ever use. One commenter emphatically declared, ‘Nobody uses the em dash in their emails or text messages. This punctuation is irrelevant to everyday use-cases.’
Oceans of communication that used to be handled by speech are now left to lone individuals typing into the internet.
This article isn’t simply a defense of dashes. Rather, I propose that the very phrase ‘everyday use-cases’ points to a profound, epoch-making transformation in our fundamental understanding of what writing is.
For a significant period in recent history, the majority of written material people consumed—beyond simple signs or menus—was what we’d call ‘formal’ writing: carefully composed, often revised and edited, and sometimes professionally printed. This type of communication stood apart from our daily interactions, which were primarily spoken. This fundamental distinction remained even with the advent of the internet.
But today, that’s no longer the case—not even close. Emails, texts, social media posts, DMs, comments, and even a DoorDasher updating you on your order—these are all vast streams of communication that once relied on speech but are now handled by individuals typing alone into the internet. Even if you love reading traditional books, you likely spend more time with these impromptu digital communications, because this is now the commonplace form of written interaction.
This casual, everyday language is truly remarkable—so creatively expressive it has even spawned a digital version of the sarcastic ‘dopey’ voice (like tYpInG LiKe tHiS). However, formal ‘writing-writing’ remains distinct. At its peak, it conveys a different kind of thought: perhaps less immediate, but often more refined, purposeful, and carefully structured, orchestrating the flow of ideas with greater precision, depth, and, yes, typically more punctuation.
Large language models are trained on immense corpora of human-produced text, including an astonishing amount of older, printed material—far more than any individual could ever process. When we ask these machines to imitate our writing, we often fail to specify—or even recognize—that our contemporary definition of ‘writing’ now encompasses the highly oral style of communication we constantly engage in via our screens. Consequently, when we review AI output and find echoes of traditional books and magazines, we perceive these familiar elements as strangely robotic. The machines are merely mirroring our historical writing conventions back to us. What we haven’t fully grasped is that we, as humans, are gradually evolving towards entirely new communication norms.
Illustration by Outlanders Design