Obvious markers of AI’: doubts raised over winner of short story prize | Books

The Crisis of Authenticity: When AI Wins the Literary Prize

The literary world is currently grappling with a digital ghost in the machine. The recent controversy surrounding “The Serpent in the Grove,” a short story that clinched a prestigious Commonwealth prize for the Caribbean, has exposed a raw nerve in the creative arts. When a work praised for its “voice of restraint and quiet authority” is suddenly flagged as a potential AI fabrication, it raises a fundamental question: Is the “human touch” still a measurable metric, or have we entered an era of indistinguishable mimicry?

The case of Jamir Nazir—a relatively unknown writer whose work was flagged by AI detectors and scrutinized by internet sleuths—is not an isolated incident. It’s a symptom of a broader shift. From the New York Times cutting ties with freelancers over AI-authored reviews to publishers cancelling debut novels due to suspected machine intervention, the industry is in a state of high alert.

Decoding the ‘AI Accent’: How to Spot Machine Prose

While LLMs (Large Language Models) are becoming more sophisticated, they often leave behind “syntactical tics”—digital fingerprints that act as a giveaway to the trained eye. Experts and critics have begun cataloging these markers, creating a shorthand for spotting generative AI.

The ‘Not X, But Y’ Trope

One of the most common markers is a specific rhythmic structure: the “not this, but that” phrasing. While humans use this for emphasis, AI tends to over-rely on it to create a false sense of nuance and sophistication.

The ‘Vocabulary of Vague Intensifiers’

Watch out for words that sound profound but lack concrete meaning. Terms like “delve,” “deeply transformative,” and “quietly powerful” are frequently used by AI to signal emotional depth without having to actually construct a complex emotional narrative.

The Profusion of Em Dashes

AI often uses em dashes to create long, flowing sentences that mimic literary complexity but often lack a clear logical progression. When a story feels “too smooth”—lacking the jagged edges of a genuine human voice—it’s often a red flag.

The Detection Arms Race: A Technical Paradox

We are currently witnessing a “continuous technical arms race.” On one side, we have detectors like Pangram attempting to isolate machine prose; on the other, we have writers using AI to “humanize” AI text, and models that are trained specifically to avoid detection.

The irony, as noted by publishers like Granta, is that AI is often the most efficient tool for detecting AI. When the AI tool Claude is asked to evaluate a piece of writing, it may equivocate—suggesting a work is neither purely human nor purely machine. This “grey zone” is where most modern literature now resides.

This creates a legal and ethical nightmare for institutions. The Commonwealth Foundation, for instance, avoids AI checkers during the judging process to protect the “artistic ownership” and consent of unpublished works. This leaves them operating on a “principle of trust”—a precarious position in an age of algorithmic deception.

The Future of Creative Provenance

As we move forward, the focus will likely shift from detection (trying to prove something is AI) to provenance (proving something is human).

We may see the rise of “process-logging,” where authors provide version histories, handwritten notes, and timestamped drafts to verify the organic growth of a story. In a world where the final output can be faked, the journey of creation becomes the only true certificate of authenticity.

the “Turing test” for literature is no longer about whether a machine can write a beautiful story—it clearly can. The real test is whether we, as readers, value the beauty of the prose more than the soul of the creator.