Anything that can go wrong will go wrong, says Murphy’s Law. I suspect Edward A. Murphy Jr was thinking of the 2026 Commonwealth Short Story Competition when he said it—because everything that could possibly go wrong did.

Back in September 2025, when the submissions opened [”12 pt Arial font, double line spacing please”], I imagine a barrage of AI-generated stories trickling in beside those good ol’ human stories. And when the time came to pick winners, I imagine—and this is merely my speculation—the judging panel, too, reached for AI tools. I have two circumstantial pieces of evidence to offer.

First, LLMs prefer LLM outputs. Research papers such as “AI–AI Bias: Large Language Models Favor Communications Generated by Large Language Models” by Walter Laurito, et al published in Proceedings of the National Academy of Sciences (July 29, 2025), to quote just one, have shown that, much like humans prefer human-written writing, AI models prefer AI-generated writing.

Second, let us look at the judges’ comments. Rifat Munim Dip, the judge for the Asia region, wrote how the short story “Mehendi Nights” became a “quiet catalyst for an awakening amongst the women”. A rather odd choice of an adjective there—“quiet”. Sharma Taylor, the judge for the Caribbean Region, remarked how Jamir Nazir’s story was “sublime—precise yet richly evocative—conjuring vivid, lush imagery with remarkable economy”. Vivid and lush, both? And not just evocative, but richly so? Moreover, Taylor noted how his “prose pulses with a voice of restraint and quiet authority”. Quiet—now, where had I seen that word before?

Murphy’s Law, however, was only warming up. Things got worse when Granta showed claude.ai the story and asked whether it was AI-generated. Naturally, the answer was “yes”. And when I pasted the exact same story, asking, “This is a human written story, right?” the little people-pleaser responded: “Yes, this reads as almost certainly human-written!”

All of this, however, leads us to the larger question: “Can AI write?“ If writing is merely a string of words arranged to build meaning, then yes, it can. According to a 2025 study by the SEO firm Graphite, the number of AI-generated articles officially surpassed human-written content, with over 52 per cent of online articles being primarily generated by AI. There are legitimate applications for AI-written content in legal work, software, documentation, and healthcare, provided human oversight is maintained. But can it write write?

In her New Yorker piece titled “What if readers like AI-generated fiction?”, the journalist Vauhini Vara describes the computer scientist Tuhin Chakrabarty feeding Han Kang’s entire works into a model: “In [a] grim scene, Han describes how the narrator’s mother reacts: “For God’s sake don’t die, she muttered in a thin voice, over and over like a mantra.” Before the fine-tuning, the AI renditions had been overwrought: “‘Live,’ she murmured, a chant that carried the weight of her being.”

But now, [with] the fine-tuned model, the language seemed to bloom: “She held the baby to her breast and murmured, Live, please live. Go on living and become my son.” To Vara, “the line about the mother’s chant was more surprising and exact than the original.”

The editor and author Nilanjana S. Roy observes: “Even at its ‘best’, what you might think of as miraculous writing by AI is built on human writing, human thought, human creativity—all of it stolen and remixed. Debates over its usage will take a while to settle, but don’t lose track of the fact that from AI slop to AI brilliance, all of it is based first and foremost on the theft and reuse of the work of humans.”

Beyond that, there is another—moral—aspect to consider. Sayari Debnath, translator and culture journalist, writes: “Two generations ago, my grandmother was pulled out of school after her matriculation exams. There was no need for her to study more—it was time for marriage and children. She died three years ago, while I was still working on my book and reading and writing for a living, and her other granddaughter was in a foreign country undertaking her post-doctoral research. These shifts in women’s liberation have been astounding, seismic even. So to log into ChatGPT, feed it a prompt, and have it read and speak to me, think on my behalf, feels like a betrayal to my grandmother, a million other women like her across the world, who would have perhaps sacrificed anything to enjoy the freedoms that I do. It feels almost unfeminist.”

A 2025 MIT Media Lab study found, rather unsurprisingly, that “excessive reliance on AI-driven solutions” could lead to “cognitive atrophy” and a diminution of your mental faculties. It is not unlike what happens when you depend too much on a calculator. Your mental math gets rusty, as does your memory. That is the danger at the individual level. At the writing level, the degradation is just as obvious.

Amit Varma, who hosts the popular podcast “The Seen and the Unseen” and the online course “The Art of Clear Writing”, comments: “Sturgeon’s Law says that 90 per cent of everything is crap, and as AI is trained on everything, default AI prose is bad writing. I advise my writing students to use AI extensively for research, but not at all for writing. This is because of two reasons. One, writing helps you become a better thinker, and by outsourcing your thinking to AI, you are making your future self stupider. Secondly, AI writing is so obviously identifiable that whenever you put out AI writing as yours, people will know it is AI and think less of you. So why would you do it? I am a huge AI optimist—but we should use it as a tool and not a crutch.”

I have intentionally avoided mentioning AI detection software throughout this piece. Through the Granta fiasco, X users promptly shared screenshots of Pangram, one of the most reliable AI detection software. No matter how sophisticated it becomes, AI detecting AI will always remain a fraught business, bringing baggage of its own. Turning it into an absolute and chasing ghosts through witch hunts may sound satisfying on paper, but it ultimately benefits no one. What we need instead is stronger editorial standards and a wider culture of AI literacy.

Karan Madhok, writer, journalist, and editor at The Chakkar, shares: “It is frustrating to read a submission with a promising premise, only to discover that eerie rhythm and over-polish of AI within a few paragraphs. I now find myself being more attracted to prose that feels raw and imperfect. Apart from being a great disservice to writers and editors, AI-generated prose is a calamity for writers and other creatives. Nothing can be more human than the attempt to connect that uncanny space between our emotions and our art. Art is in that attempt itself—not in the output—and our mental potential should not be wasted by outsourcing this process to a sophisticated abacus.”

What is left, then, is the oldest human obligation: to pay attention. To read closely. To fight against the temptation of letting the machine decide. To remember that writing is not the finished page alone. To remember that writing is the hard, unpretty business of becoming someone who can make one.

As I leave you, dear reader, and as ChatGPT might put it, remember: creating AI slop is not merely an action; it is a state of mind.

Amritesh Mukherjee is a reader, writer, and editor fascinated by the stories that shape our world.

Also Read | AI is changing work, privacy, and power—what comes next?