What started as a lighthearted experiment by British fantasy author Mark Lawrence ended with an unsettling twist: AI appears to write better short stories than humans—at least according to the readers. Lawrence invited four award-winning fantasy authors, including himself, to each write a 350-word story about “a demon.” He gave the same prompt to GPT-5 (ChatGPT’s latest version), with slight tonal or setting tweaks to encourage variation. Eight anonymous stories—four human, four AI—were posted on his blog, and fans were asked to vote for the best and guess the authorship.

The result: not only did the top-rated story come from AI, but on average, the AI stories scored higher than those written by humans. Readers failed to distinguish machine from man—suggesting a literary Turing test may have been passed.

Lawrence, though fascinated, expressed deep concern. “I hate that AI can do this,” he wrote, calling the outcome a “huge shock.” He noted that AI-generated content exists only because it was trained on works by human authors, often without their consent.

Although AI still struggles to maintain coherence over novel-length fiction, the experiment raises fresh questions about the future of creative professions. How should society respond when synthetic creativity rivals the original?