July 5, 2026 – A new wave of suspicion is sweeping through the literary and media worlds, fueled by a single, unsettling reality: most people cannot reliably tell the difference between human writing and text generated by large language models (LLMs). As of this weekend, the debate has reached a fever pitch after a debut horror novel was abruptly pulled by Hachette following rumors of AI involvement, and a prizewinning short story faced intense online scrutiny in May.
Forensic linguists warn that the public’s confidence in spotting AI prose is dangerously misplaced. Claire Hardaker, a professor at Lancaster University, has developed an online test called “Bot or Not” that asks users to identify fake hotel reviews. Her data shows that even trained readers get it right only about 60% of the time. “People have learned very simplistic rubrics,” Hardaker told The Guardian, “and now just madly apply them everywhere.” She points to common markers like the “rule of three” or excessive use of dashes, but notes these are also hallmarks of classic human writing—from Charles Dickens to Julius Caesar.
The fallout is immediate. In May, author Jamir Nazir faced a firestorm of allegations after winning a prize for a short story, with social media users claiming they could “just tell” it was AI-generated. Nazir later insisted to The Atlantic that he did not use the technology. Yet the damage was done. Now, with the withdrawal of the novel “Shy Girl,” publishers are scrambling for new verification protocols. The problem is that LLMs are trained on vast datasets of human text, making their output statistically indistinguishable from a competent human writer’s first draft.
This ambiguity has created a chilling effect in creative circles. Novelists like Jennifer Egan and Jeanette Winterson are openly debating whether the future of fiction will require watermarking or even a return to handwritten manuscripts. For the average reader, the takeaway is stark: the next time you skim a review, a news article, or even a prize-winning story, you may be guessing—not judging. As Hardaker put it, “The machines are already among us. We just don’t know how to spot them yet.”