By – Dr Srabani Basu
Associate Professor, Department of Literature and Languages , SRM University AP( Amaravati)
There was a time when an em dash could walk peacefully through a sentence. Nobody stopped it. Nobody photographed it. Nobody whispered, “Suspicious.” Writers welcomed it into paragraphs without demanding identification documents. Editors occasionally removed one or added another, but these were ordinary occupational hazards. Then GPT arrived. It discovered the em dash, apparently fell passionately in love with it, and began scattering the poor creature across prose with such enthusiasm that somewhere between 2023 and 2026, one of the oldest punctuation marks in English acquired a new identity: Exhibit A in the Case of Humanity vs Artificial Intelligence.
Today, place an em dash in a LinkedIn article and somebody will raise an eyebrow. Place two and suspicion deepens. Three em dashes and the Cyber Crime Department may be informed. Four and you might as well surrender your laptop, confess that you have never had an original thought, and wait patiently for the AI detector to pronounce sentence. The em dash, once an elegant device for interruption, amplification and dramatic pause, has become the punctuation equivalent of being found at an airport carrying an unlabelled suitcase.
Writers have therefore developed an extraordinary new literary genre: The Disclaimer. “This article was entirely written by me.” “No AI was used in writing this piece.” And now, increasingly, we need another footnote: “The em dashes appearing in this article are naturally occurring punctuation marks and are not the offspring of artificial intelligence. No language model was involved in their conception, gestation or delivery.” Somewhere, the semicolon is watching nervously. It knows fashions change.
The irony is magnificent because the em dash did not arrive with ChatGPT. It has been roaming literature for centuries. Emily Dickinson practically offered it permanent residency. Virginia Woolf knew what to do with interruption. James Joyce was hardly waiting for a chatbot to give him permission to misbehave with punctuation. Laurence Sterne was playing typographical games when artificial intelligence would have sounded like the title of a particularly implausible sermon. Writers have always stretched punctuation because thought itself does not travel obediently from capital letter to full stop. Thought interrupts itself, doubles back, remembers something inconvenient halfway through, wanders into the kitchen, forgets why it went there and returns carrying a metaphor. The em dash understands this.
But now the poor thing has a reputation.
Imagine its predicament. The em dash wakes one morning, opens LinkedIn and discovers thousands of posts explaining that “excessive use of em dashes is a sign of AI-generated writing.” It is devastated. “Excessive?” it cries. “I was expressive!” The comma avoids eye contact. The colon pretends to be busy introducing a list. The semicolon, who has spent centuries being misunderstood anyway, offers quiet solidarity. The full stop remains smug. Nobody suspects the full stop. It has always had establishment connections.
Eventually, the em dash hires a lawyer.
“Your Honour,” begins Counsel for the Defence, “my client existed long before generative AI.”
The prosecution rises. “But GPT uses your client frequently.”
“So do humans.”
“Precisely. Suspicious.”
The courtroom falls silent.
This is where the joke stops being entirely a joke.
We are entering a curious cultural moment in which human beings are beginning to abandon perfectly legitimate features of human expression because machines have learned to reproduce them. Think about the philosophical absurdity of that. AI models were trained on enormous quantities of human-created language. They learned patterns from us. Now, because machines reproduce some of those patterns effectively, humans are being told that using those same patterns makes us look artificial.
The student has copied the teacher so successfully that the teacher is now being accused of copying the student.
Welcome to the age of reverse imitation.
The consequences are already visible in the strange anxiety surrounding AI detection. A person writes an original paragraph. A detector announces that it is 78 percent AI-generated. The writer protests, “But I wrote it.” The software, presumably stroking an imaginary beard, replies, “That is exactly what AI would say.”
At this point we have left computer science and entered Kafka.
What exactly makes writing human? Is it the absence of an em dash? Short sentences? Grammatical mistakes? A typo after every third paragraph? Must I now deliberately misspell “consciousness” to establish biological authenticity? Perhaps writers will soon insert random errors as digital fingerprints. “I definatelybeleive this.” Detector: Human! 99.7% confidence. Congratulations. We have reached a civilisation in which incompetence may become evidence of authenticity.
The problem grows deeper when style itself becomes suspect. Suppose a writer naturally enjoys long, rhythmic sentences. Suspicious. Balanced contrasts? AI. Metaphors? Probably AI. Clear structure? Definitely AI. Words such as “delve,” “nuanced,” “multifaceted” and “landscape” have already acquired the linguistic equivalent of police records. Soon we may need an endangered-words programme. “Nuanced” will be relocated to a protected dictionary where it can reproduce safely without being accused of prompt engineering.
Meanwhile, the em dash sits in witness protection under the assumed identity of a comma.
This matters because language is not merely a delivery mechanism for information. Style carries identity. Some writers use fragments. Some adore parentheses. Some construct sentences that wander magnificently across half a page before reluctantly agreeing to terminate. Some like semicolons; others behave as though semicolons were invented by the tax department. Punctuation is part of rhythm, and rhythm is part of voice. When writers begin modifying their natural voices not because the writing demands it but because they fear appearing machine-generated, AI has influenced human writing in a far stranger way than simple content generation.
We are not merely asking machines to write like us.
We are beginning to write differently so that nobody thinks we write like machines that learned to write like us.
Try drawing that family tree.
This is why the current obsession with identifying AI through stylistic fingerprints deserves more critical thought. Of course, questions of academic integrity, disclosure and authorship matter. Passing machine-generated work off as one’s own in contexts where original authorship is required raises legitimate ethical concerns. But punctuation cannot carry the burden of forensic evidence. Neither can vocabulary, sentence length or a suspiciously well-behaved paragraph. An em dash proves approximately one thing: somebody, or something, used an em dash.
Perhaps we are asking the wrong question. Instead of obsessively asking, “Was AI involved?”, we may sometimes need to ask, “Where is the thinking?” Does the writer have something to say? Is there an argument? A perspective? Intellectual risk? Lived observation? A surprising connection? Can the person defend the idea, extend it, question it and explain how they arrived there? Those questions take more effort than feeding prose into a detector, but thinking has always been inconveniently resistant to automation.
AI literacy should therefore involve more than learning how to use AI. It should also involve learning how not to become intellectually paranoid about it. Tools change. Writing changes. Conventions change. Human creativity has survived the printing press, typewriter, spellcheck, predictive text, Grammarly, search engines and the catastrophic invention of autocorrect. It will probably survive ChatGPT too. But only if we resist reducing authorship to a collection of superficial stylistic clues.
So perhaps it is time to rehabilitate the em dash.
Let it return to our sentences without producing panic. Let semicolons emerge from their bunkers. Let “nuanced” apply for parole. Let writers use long sentences if their thoughts require long sentences and short ones when brevity bites harder. Let punctuation serve thought rather than digital suspicion.
And the next time somebody points accusingly at an em dash and declares, “Aha! AI!”, perhaps the em dash should finally turn around.
It has endured enough.
It will straighten its tiny horizontal spine, summon several centuries of literary history and say:
“Excuse me. AI learned me from you.”
Then it will produce its birth certificate, call Emily Dickinson as a character witness, and file a defamation suit.
The semicolon has already agreed to testify.
The comma, unfortunately, cannot attend.
It has separation anxiety.
