Mary Shelley Was a Fraud
AI detectors accuse a nineteenth-century literary masterpiece of being AI-generated, fuelled by synthetic authority, corporate arrogance, and the mass production of technological slop.
“Mary Shelley was a fraud” sounds outrageous, right? Well, it sounds outrageous until you realise some AI detectors effectively made that argument themselves. Oh, the insanity!
Having written extensively about AI, LLMs, and the growing collapse of trust surrounding both, I felt compelled to weigh in on this discussion. Over the last few weeks, social media users began feeding passages from Shelley’s Frankenstein into AI detection programs including GPTZero, Copyleaks, Scribbr, and other widely used tools now embedded throughout academia and publishing. In several circulating examples, passages from Shelley’s 1818 novel received scores suggesting partial or substantial AI authorship. A novel written more than two hundred years before the invention of LLMs triggered warnings for AI. Huh?
People treated the result as a joke, Reddit threads exploded with disbelief, online forums mocked the detectors for accusing one of history’s most studied literary works of sounding algorithmic, while others defended the software and argued that old writing styles confuse modern detection systems.
That defence makes the situation worse.
If a detector struggles to distinguish between canonical human literature and machine output, then why are universities, publishers, literary agencies, editors, and employers treating these systems as legitimate arbiters of truth? The whole system starts to feel less like technological progress and more like an industry protecting itself with smoke and mirrors, whilst quietly extracting value from fear, uncertainty, and institutional dependence.
And that question is the crux of this whole debate. Yet nobody inside institutional power structures wants to answer it. Or, at the very least, answer it honestly.
The Mary Shelley example carries significant weight because it exposes the central deception behind AI detection tech. These systems don’t understand writing, they don’t understand originality, and they clearly don’t understand literature, voice, intent, symbolism, ambiguity, emotion, satire, rhythm, experimentation, or historical context. They identify statistical familiarity. That’s it.
I wrote about this in my August 2025 essay on Medium, which was subsequently published on a major AI platform, titled “When a Full Stop Becomes AI: Questioning the Reliability of AI Detection Tools”.
AI Detectors?
What do these programs do exactly? Well, the detector scans sentence structures (yes, plural), punctuation patterns, vocabulary distributions, predictability scores, and stylistic consistency. Then it compares those traits against probability models trained on existing human writing and machine-generated text. Which means the entire system is built on a contradiction from the ground up.
AI learned to imitate human language by consuming colossal amounts of human language. Then Silicon Valley, aka BigTech, built detectors to identify writing that resembles the language patterns their own systems absorbed from writers in the first place. Yep, writers trained the machine, and now the machine accuses writers of sounding like itself. That’s nothing short of cultural cannibalism wrapped in corporate branding.
Meanwhile, the flying monkeys, or defenders, of AI detection software continue repeating the same exhausted argument: the tools are imperfect but useful. During the COVID era, for example, we all heard nearly identical language surrounding fact-checking systems embedded across media platforms and social networks. Those systems arrived carrying the same promise of institutional protection (think Fact Checkers, Snopes, PolitiFact, et al). Trust the experts they told us. Trust the system. Trust the labels appearing beneath content.
Years later, we all woke to the scam that took place and recognised what had happened. Fact-checking operations frequently operated less as neutral arbiters of truth and more as enforcement mechanisms for approved narratives. Major stories dismissed as misinformation (aka conspiracy theories) later turned out to hold substance and, dare I say it, come to fruition. Public trust collapsed because institutional certainty repeatedly collided with reality.
AI detectors now occupy similar territory. The detector produces a score. The institution treats the score as authority. The accused writer must defend their humanity. That inversion alone should alarm every serious writer alive.
Across universities worldwide, students increasingly face accusations based on software outputs that even the companies behind the systems admit are unreliable. Yet institutions continue deploying them because they provide something modern bureaucracies value more than truth: plausible deniability.
A professor no longer needs to explain why a paper feels suspicious. An editor no longer needs to articulate literary concerns. A publisher no longer needs to risk investing time in reading difficult work. Because the software flagged it. Case closed. It’s intellectual laziness disguised as technological sophistication, embraced by institutions that increasingly prefer automation over thought, risk avoidance over judgment, and convenient software outputs over the hard work of reading, analysing, and thinking for themselves.
And this is precisely where the Mary Shelley debate becomes frighteningly devastating.
Frankenstein didn’t confuse the detectors because Shelley wrote badly. It confused them because LLMs increasingly mimic the polished, grammatically coherent, structurally balanced prose common to educated writing. Writers developed those conventions over centuries. AI absorbed them through mass ingestion and then regurgitated them as a dense, unstable mix of borrowed patterns stripped of context, intent, and authorship. That’s what remains when they’re all stripped away. Pure slop.
Then – and here’s the irony - the AI detector industry arrived and reframed those same conventions as “suspicious”. People’s obsession with em dashes offers one of the clearest examples of this insanity. At one bizarre point across the social media spectrum, people were accusing writers of using AI because they were employing em dashes, clean syntax, or orderly paragraph structures. Some writers even started altering their natural voice to avoid false accusations. Think about the absurdity of that cultural shift - writers modifying their styles to avoid resembling machines that learned language patterns from… writers!
Nobody inside Silicon Valley seems interested in discussing the psychological implications of that loop. Then again, tech companies aren’t necessarily built by great independent thinkers. They’re often built by highly skilled coders, engineers, and programmers operating inside systems that reward optimisation, scale, conformity, and profit far more than reflection, wisdom, or intellectual depth.
But wait, there’s more. The contradiction becomes even uglier once copyright enters the conversation.
In August 2024, authors filed lawsuits against Anthropic and other AI companies alleging unlawful use of copyrighted books to train LLMs. Writers argued that corporations harvested human intellectual labour on an industrial scale without permission or compensation. At the same time, those same corporations and affiliated industries increasingly promoted systems that accused writers of sounding artificially generated. Welcome to The Twilight Zone, folks.
And don’t think for a minute that this is part of “progress”, because the implications are staggering. An author spends decades developing a voice. A machine consumes millions of books containing voices like theirs in days. The machine reproduces statistical echoes of those patterns, and then identifies the original (human) style as suspicious because the AI model already absorbed it.
So what the hell exactly are writers supposed to do with that? Write worse? Write messier? Insert deliberate grammatical flaws to prove humanity? It breaks my heart to say this, but many already do. What we’re actually witnessing is a slow collapse of confidence in human intellect in real time.
University lecturers and professors – those woke career academics with doctorates in gender politics who come from a breeding ground for mediocrity where outrage, conformity, and a lack of self-awareness thrive — openly advise students to avoid “overly polished” writing because detectors might flag them. Sure, let’s allow a failed education system rewrite the rules of language. Online communities now exchange tactics for “humanising” prose. Some writers intentionally break sentence flow, remove punctuation, introduce awkward phrasing, or simplify vocabulary to reduce suspicion. I call that cultural regression driven by technological paranoia.
Meanwhile, companies behind AI detectors continue marketing certainty they can’t reliably provide. GPT Zero markets itself heavily within education. Copyleaks promotes AI detection for institutions and enterprises. Scribbr presents its detector as an academic integrity solution. The branding language always sounds authoritative — precision, accuracy, confidence, trust.
Yet independent testing repeatedly reveals contradictions, false positives, inconsistent results, and detector disagreement across identical texts. One detector flags a passage as human. Another flags it as AI. A third produces mixed probability scores.
The weak institutions blindly relying on these systems rarely discuss those inconsistencies publicly because the illusion of authority matters more than the reliability of the result. That pattern appears everywhere in modern tech culture. Build first, scale aggressively, capture institutional dependence, and normalise flaws later.
People defending these systems often respond emotionally when criticism appears. Anyone questioning AI detectors gets labelled anti-tech or resistant to progress. Yet criticism of unreliable systems is not anti-tech, it’s the foundation of intellectual seriousness. And as we all know, technology without scrutiny becomes ideology.
The Shelley debate spread online because it compressed the entire problem into one unforgettable image: a literary masterpiece from 1818 accused of being artificial by software built in the twenty-first century.
Frankenstein now feels disturbingly prophetic in ways Shelley herself never intended. Her novel explored humanity’s obsession with creation detached from moral responsibility. Victor Frankenstein became consumed with the act of building life while refusing accountability for what emerged afterward.
Two centuries later, Big Tech operates with similar psychology: build the system, release the system, scale the system, monetise the system, and deal with the consequences later, while directing enormous amounts of energy into defending the technology, protecting the brand, and ridiculing the very writers whose work made the system possible in the first place.
The public already sees the damage spreading through education, journalism, publishing, art, music, and online communication. AI-generated slop floods search engines and social feeds while authentic (human) work struggles for visibility inside algorithmic ecosystems designed around engagement volume rather than intellectual value.
Now the same tech culture responsible for flooding the internet with synthetic language sells detection systems claiming to protect humanity from synthetic language. It’s akin to the arsonist arriving carrying a fire extinguisher.
That’s why the Mary Shelley debate resonated so strongly online. Because people instinctively recognised the absurdity, and because the absurdity exposed the larger truth.
AI detectors are dressed up as guardians of originality while doing little more than guessing patterns and calling it authority.
And before the angry mob comes at me with pitchforks and torches screaming the “false positive” rhetoric, the danger extends beyond false positives. The deeper threat involves what these systems train society to believe about creativity itself. Once institutions accept algorithmic probability scores as substitutes for human judgment, originality gets reduced to something a piece of software is asked to approve before anyone is allowed to read it.
That shift corrodes culture at its foundation because great writing breaks patterns while machines reward them, and that clash was always going to happen.
Mary Shelley exposed it simply by existing within the system being tested.
This essay is a polemic shaped by years spent evaluating AI systems, LLMs, detection software, online discourse, and the broader cultural effects surrounding them. The arguments presented here emerge from ongoing research, independent analysis, and a body of previously published essays examining the growing collision between technology, authorship, institutional trust, and human creativity.
PCR testing came to mind while reading this.
So, Shelley's text was flagged because these same classic books had been used to train AI on syntax and AI will now generate something very similar?