It’s 2025, and I find myself caught in an ironic 21st-century witch hunt. Only this time, the torches and pitchforks have been replaced by AI detectors and self-appointed “sleuths” sniffing out supposed machine-written prose. After months of hearing horror stories from fellow authors and academics I found my audiobook in the crosshairs of being deemed AI. So apparantly my voice over actor, Jan Lanham, is a fucking robot? I’m pissed. I am fucking pissed that I had to defend my work and my voice over actor to get my audiobook relisted on a platform.
A fellow indie writer uses em dashes frequently—that darling of punctuation—and suddenly the Twitterati (sorry, X-ites) screamed “ChatGPT wrote that!” so loudly that she felt she had to leave the platform. An awkwardly formal paragraph? Clearly the work of a soulless algorithm, they insist. The irony is rich as fuck: in an era when authors struggle for authenticity and voice, we’re getting accused of not being human. Welcome to the literary witch hunt of 2025.
History Repeats: From Typewriters to Chatbots
This moral panic isn’t new. The publishing world has always feared the latest tool would corrupt writing. Go back a few decades when word processors first showed up, many authors recoiled like the little snob bitches they are.
In the 1980s, British writers sneered that “processing was for peas, not words.” They fretted that a computer had a “mind of its own” which might impose “android habits” on their prose. Some even claimed the uniform look of word-processed text made every manuscript seem like it sprang “from one giant brain.” This is a hilariously paranoid notion in hindsight. But it’s a pattern almost as old as the written word.
Quills vs. fountain pens, pens vs. typewriters, typewriters vs. word processors! Each new writing tool sparked hand-wringing in its day. Today’s pearl-clutching witch hunt over AI in writing is just the latest verse in this repetitive ballad of angst. And it is not even the cool punk angst, just ignorant bullshit that hurts people!
The difference now is that our modern inquisitors come armed with algorithmic “evidence.” Yesterday’s skeptics only feared the typewriter or word processor; today’s sleuths think they can prove you used ChatGPT because you used an uncommon turn of phrase. It’s as if the old guard who once mistrusted WordStar and WordPerfect have been reborn as digital detectives, zealously hunting for any spectral hint of AI possession in our sentences.
The False Positives Problem
Here’s the rub. The tools these sleuths rely on, that often use AI, are about as reliable as a dowsing rod in a fucking desert. A cottage industry of AI-text detectors has popped up. Turnitin’s AI checker for schools, the popular GPTZero, online tools like ZeroGPT, and others. They boast impressive-sounding accuracy rates, but reality tells a different story. Even OpenAI (yes, the folks behind ChatGPT) shut down its own AI-writing detector for being too inaccurate.
Turnitin, the academic plagiarism giant, admitted that its AI detection tool incorrectly flags about 4% of human writing as AI on a sentence-by-sentence level. That sounds reasonable and awesome, right? Four percent of 70 million scanned assignments is a lot of innocent students falsely accused. Just shy of three million students and academics falsely accused. And those are the vendor’s own numbers, given with a straight face.
Independent tests make the picture look even worse. In one study, the ZeroGPT detector misidentified 83% of human-written research abstracts as AI-generated. Another research team found a detector labeling over 60% of genuine academic papers and essays as AI-written. Think about that for a moment. These supposed truth machines were dead wrong in multiple peer reviewed studies more than half the time.
These tools cry “AI wolf!” at authentic work. No wonder Reddit and Threads student forums are now rife with horror stories of false positives. The Washington Post’s tech desk started hearing from “many angry high school and college students (and some of their parents) claiming they had been falsely accused of AI cheating” after schools unleashed detectors like Turnitin and GPTZero. Careers and academic records are at stake, all because some algorithm had a hunch. What is an acceptable error ratio when we are talking about academic careers or student GPA’s?
The false positive fiascos can get downright absurd. One popular detector, GPTZero, even claimed that the United States Constitution was written by AI. Not to be outdone, another tool said hold my beer and declared an excerpt of Jane Austen’s Pride and Prejudice to be 73% AI-generated. Jane Fucking Austen! It would be hilarious if it weren’t so indicative of a broken system. In that same test, one detector confidently screamed “AI” at Austen’s prose, while another detector judged it 100%. When tools can’t even agree on Jane Austen, how are we trusting them to judge students and writers? These systems flip-flop more than a politician at a press conference.
The companies behind these detectors plead that false positives are rare. Originality.AI, for instance, claims only 1.5% of human texts get wrongly flagged. Copyleaks boasts a minuscule 0.2% false positive rate. And ZeroGPT crows about “98.8% accuracy.” But those boasts crumble under scrutiny. In practice many detectors are far from reliable. Multiple peer-reviewed studies have basically torpedoed the notion that AI detection is anywhere near foolproof, especially on longer-form and creative writing. Longer-form creative writing is a fancy term for books. Like my book that I had to defend from an online retailer (and did). It turns out that writing is a complex, rich human activity that is not easily reducible to the simplistic statistical fingerprints these tools look for.
Em Dash Inquisitors and Other Absurd Clues
Speaking of easily reducible simplistic fingerprints… let’s talk about the tooless tools that are the self-styled AI sleuths that step in to fill the paranoia void. Lately, these fuckers have been zeroing in on… punctuation. The humble em dash (—) has been derided by internet commenters as the telltale “ChatGPT hyphen,” supposedly a punctuation mark overused by AI text.
In early 2025, would-be detectives on social media began insisting that an em dash in writing was evidence of bot witchcraft that must be burned at the literary stake. “Em dashes are relatively rare when a human uses it… But AI chats love using it,” one amateur sleuth proclaimed confidently in a Washington Post article. This is really piss poor shitty punctuation phrenology and about as scientific as accusing someone of witchcraft because she had a black cat.
Writers and editors and college professors have been quick to push back on the em dash hysteria. The em dash is a beloved tool of human expression – from Emily Dickinson’s poetry to modern journalism – and plenty of us flesh-and-blood scribes use it liberally. “The idea that it is an indicator of soulless, dead AI-generated writing is really upsetting to me,” said one journalism professor, noting she and her colleagues use em dashes all the time.
Yet such is the sad and pathetic desperation to find some silver-bullet sign of AI text that even punctuation has become a suspect. When did we arrive at a point where using proper punctuation marks makes the bastards think a bot wrote your content? It is a catch 22. Make a grammar error and the grammar nazis come after you. Compose it well and the AI squad comes at you. Clowns to the left of me and jokers to the right…
The em dash isn’t the only victim of this absurd hunt for “AI tells.” Certain words have been branded suspicious as well! If you “delve” into a topic, for example, some folks will arch an eyebrow and claim delve is an “AI buzzword” that real humans supposedly never say even though I just said it (Damn! AM I a replicant???). It’s a witch hunt mentality, seeing spells and sorcery in every stylistic flourish. A thoughtful, slightly formal tone gets tagged as “stilted so it must be AI.” A coherent, structured essay gets accused of being “too organized, must be machine-made.” It would be comical if it weren’t hammering real writers’ reputations.
These so-called sleuths are chasing ghosts in the machine. In centuries past, inquisitors saw the Devil’s hand in every misfortune, today, a subset of editors and readers see OpenAI’s hand in every em dash or SAT word.
The truth is, there is no single gimmick or quirk that definitively marks a piece of writing as AI-made. AI text generators mimic human style quite capably, and humans can coincidentally write in ways that trigger an AI detector’s pattern-matching. Even the AI-detection companies admit their craft is a “tricky science” at best, with imperfect tools prone to false positives. The hunt for a surefire telltale sign is a wild goose chase and it’s leading to a lot of wild accusations that are hurting students, academics, and writers.
Hard Times for Real Writers
For those of us who actually put pen to paper (or fingers to keyboard) for a living, this climate is beyond frustrating. It’s hard enough being an indie author or journalist today with tight budgets, crowded markets, and readers conditioned to expect free content (I am sure some of you bitched about the paywalls in some of my hyperlinks. Sorrynotsorry that reporters at WAPO deserve an income).
Now we have to fear that if our prose is too eloquent, or heaven forbid we use a semicolon correctly, some armchair AI cop will cry “J’accuse!” and try to discredit our work. It’s a fucking Kafkaesque scenario: Write too poorly and nobody reads you; write too well and nobody believes you actually wrote it. Damned if you do, damned if you don’t.
I’ve heard from writer friends who got flagged by clients or editors simply because their draft “felt” AI-ish. One was told an internal AI checker flagged their article at 40% likely AI. The smoking gun? Her use of long, complex sentences. Students face similar perils! A perfectly innocent essay that’s a bit polished can trip the alarm. Imagine pouring your heart into an assignment or story, only to be told a computer the prof used thinks you cheated. In academia, some professors are trusting these tools implicitly. You are effectively guilty until proven innocent by a robot’s say-so that you didn’t use a robot? Is this where we are?
The personal toll is real. Stress, wasted time, and damaged reputations all thanks to false accusations. Writers have been forced to defend their humanity, compiling drafts, notes, and revision histories as evidence that “I am not a witch…I mean… machine.” Some have resorted to deliberately dumbing down their style or adding typos to avoid seeming “too perfect” (how’s that for perverse incentive?)
On a forum, a user lamented how AI detectors “penalize good writing because it looks too much like how AI generates text,” effectively punishing “the writing habits we learned in school.” We’re at a point where writing well can make others suspicious of you? It’s insane! An experienced editor once could instantly recognize a writer’s voice in a manuscript. Now a clueless algorithm might flag that voice as “non-human” because it’s consistent and polished.
And let’s not forget the resources wasted on this witch hunt. Schools, universities, publishing companies, and online retailers are spending money on AI-detection software subscriptions, training sessions about how to interpret the dubious reports these tools generate, on disciplinary proceedings sparked by nothing more than an already provably flawed AI score. This is time and money not spent on mentoring writers, improving media literacy, or, I dunno, paying authors fair wages. Indie publishers and content creators with scant resources are especially hurt. They can’t fight back easily against a false accusation like I did today, nor can they afford fancy tools to “prove” a negative. The playing field, already uneven, tilts further against the indie writer.
Ending the AI Witch Hunt
It’s high time we point a spotlight at this new literary inquisition and call it what it is… a paranoid overreaction. Yes, AI text generators exist, and yes, some people abuse them. But deploying unreliable detectors en masse, or turning every editor into an AI detective scouring for secret tells, is not the answer. We’re setting up a dystopia where writers have to write for the algorithm, contorting our style to avoid false flags. That’s death to creativity and voice. It’s the equivalent of those medieval witch trials where if you drowned, you were innocent.
Instead of witch hunts, we need a level-headed approach. Trust but verify, as the saying goes. If a text seems suspiciously beyond a writer’s known ability, have a conversation. Ask them about their process, maybe even have them write a paragraph in person. Remember that detectors can be duped and defeated, often by other AIs or simple rewrites. In the end, nothing replaces the discernment of a human reader who knows context, voice, and the limits of these tools (limits that are swiftly reducing).
Most importantly, let’s dial down the hysteria. The publishing world survived the typewriter. It weathered the word processor and the spell-checker and the autocorrect. Writers still wrote. Readers still read. Today’s AI writing models are just another tool. Powerful? Yes! It likely requires adaptation, but not the end of writing as we know it. We don’t need a witch hunt, we need a conversation. We need updated ethics, transparency from authors about what tools (AI or otherwise) they use, and maybe new norms for certain kinds of writing. What we don’t need is a bunch of half-baked algorithms and hyperventilating sleuths turning the literary landscape into Salem circa 1692.
So to the AI detectives and witch hunters, I say: take a deep fucking breath. Not every well-placed em dash is an invocation of dark magic. Sometimes an em dash is just an em dash — a handy punctuation mark that we humans like to use. Let writers write. By all means, we should not tolerate those who cheat and plagiarize, but don’t burn the rest of us at the stake to do it.
The only thing these AI witch hunts are guaranteed to destroy is our trust in each other’s words. And for a community built on words, that trust is everything. We’ve been warned before in past panics, cooler heads eventually prevailed, but there has always been a cost first. Let’s learn from that. It’s time to end the witch hunt and get back to the business of writing. Messy, human, beautifully imperfect writing without an algorithm or jag-off peering suspiciously over our shoulders.
Stay totally awesome!
Stay true to you!
Still Reading?
Hey… thanks for sticking around to the end.
If this piece hit a chord or maybe made you think, I’d be honored if you’d check out my book, Hearts of Glass Living in the Real World. (by me- Pat Green)
It’s about survival, love, and the kind of found family that keeps you standing when the world gets mean. You can find it in print, ebook, or audiobook pretty much anywhere you get those things online. If you have trouble finding it you can get it directly from my publisher, Barnstormer Publishing! We can even hook you up with a signed copy for a few bucks more!
If you pick it up, thank you. If you review it, even better. Either way, just know this: your time, your support, your voice… they all matter more than you know.
Leave a Reply