When trying to spot if a piece of writing was crafted by artificial intelligence, many look for subtle clues. But recently, the internet became obsessed with a single punctuation mark: the em dash. ChatGPT, it turned out, was absolutely smitten with it—using it everywhere, even when explicitly told not to.
This peculiar observation spread rapidly, leading to a strange internet-wide agreement: that humans simply don’t use dashes. In online tech communities, these dashes were quickly labeled a “GPT-ism,” an impersonal, robotic quirk that supposedly doesn’t fit modern communication styles. One user on an OpenAI forum even lamented that the frequent dashes made it challenging to use ChatGPT for customer service without clients realizing they were interacting with a bot. It seemed everyone was convinced that no real person would ever employ such punctuation, and anyone who dared to would be instantly flagged as an AI.
Naturally, these ‘deviant’ dash-users were horrified—and I count myself among them. As a former proofreader, I could passionately discuss the subtle nuances of the even narrower en dash for hours. While I admit my enthusiastic dash usage might be uncommon, I never imagined its very existence would be challenged. The dash is a venerable and perfectly conventional tool for crafting sentences! Just look at the works of Dickens, Dickinson, Nietzsche, Stephen King, or even this very magazine—they’re all brimming with dashes. Indeed, part of their appeal is their ability to convey a more casual, human-like cadence, mirroring natural speech better than strict colons, semicolons, or parentheses. After all, humans don’t always think or speak in perfectly formed sentences; we communicate in streams of thought, which often interrupt, introduce, and intertwine in a fluid dance of complex ideas. (Sometimes, this ‘dance’ isn’t so neat: Salinger’s dialogue, for example, wonderfully illustrates how frequently our thoughts are cut short or left incomplete.) This fluid expression, in essence, is what punctuation is for.
Interestingly, the most telling AI ‘tell’ isn’t the dash itself, but its orthography—how it’s styled. ChatGPT meticulously renders its em dashes in the classic print-book fashion: a line the exact width of an ‘M,’ with no spaces around it. Most everyday computer users simply don’t type this way. They might not even know the specific key combination required to produce this character (or its proper name; some online chatter referred to it as a “ChatGPT hyphen”). Instead, the typical user will opt for a single hyphen (-) or a double hyphen (–), which some programs might auto-correct to the less-common en dash (–). Crucially, human users almost always add spaces around their dashes, a common practice in online publishing that aids readability and ensures cleaner text wrapping.
Despite these technical distinctions, the online debate stubbornly fixated on the dash itself. People discussed it as if it were a bizarre, archaic symbol that no modern, self-respecting human would ever use. One commenter emphatically stated, “Nobody uses the em dash in their emails or text messages. This punctuation is irrelevant to everyday use-cases.”
Oceans of communication that once relied on spoken words are now solely handled by individuals typing on the internet.
My aim here isn’t to defend the humble dash. Rather, I want to argue that the phrase “everyday use-cases” points to a truly monumental shift in our fundamental understanding of what writing actually is.
Think about it: for a significant period in recent history, most written material that people regularly encountered—anything beyond simple signs or menus—was considered ‘proper’ writing. This meant text that someone carefully composed, often revised, edited, and perhaps even professionally printed. This form of communication stood apart from our daily interactions, which were predominantly spoken. Even with the advent of the internet, this underlying distinction largely endured.
Today, however, that distinction has all but vanished. “Emails or text messages,” social media posts, direct messages, chat conversations, even a DoorDasher informing you about an out-of-stock item—vast amounts of communication once reserved for speech are now carried out by individuals tapping away on their devices. Even if you consider yourself a diligent reader, you likely spend more time engaging with quick, informal typed messages, because that has become the prevalent form of everyday writing.
This casual, everyday language is undeniably wonderful—so creatively expressive that it’s even birthed its own typographic equivalent of a mocking tone. (Think ‘tYpInG LiKe tHiS.’) Yet, formal ‘writing-writing’ remains distinct, doesn’t it? At its peak, it captures a different level of thought: perhaps less raw and immediate, but frequently more refined and intentional, meticulously structured to orchestrate the flow of ideas with greater precision, depth and, yes, careful punctuation.
Generative AI models learn from an immense corpus of human-produced text, including a vast amount of historical printed material that far exceeds what any single person could ever consume. We instruct these machines to emulate our writing, but we often fail to—or don’t even recognize—that our current definition of ‘writing’ now encompasses the highly oral style of communication we constantly send through our digital devices. When we then scrutinize AI’s output and find echoes of old books and magazines, we wrongly label these artifacts as ‘robotic.’ In reality, these machines are merely reflecting our own rich traditions back to us. What we might not yet grasp is that we are, in fact, evolving towards entirely new traditions of written expression ourselves.
Illustration: Outlanders Design