.jpg)
What the Study Reveals About AI Writing
In December 2025, researchers at University College Cork (UCC) published a groundbreaking paper that challenges the notion that large language models can replicate human creative writing. Using literary stylometry—a technique traditionally applied to authorship attribution—the study compared hundreds of short stories written by humans with those generated by AI systems such as GPT‑3.5, GPT‑4, and Llama‑70B.
Methodology: Literary Stylometry in Action
Dr James O’Sullivan and his team extracted subtle linguistic markers from each text, including word‑frequency distributions, sentence‑length variability, and rhythmic patterns. By applying clustering algorithms and multidimensional scaling (MDS), they visualised the stylistic space occupied by each author and each AI model.
Key Findings: Stylistic Fingerprints
- AI‑generated prose consistently clusters into tight, predictable groups, reflecting the uniformity of the underlying model.
- Human writers exhibit a broader range of stylistic choices, driven by personal voice, creative intent, and lived experience.
- Even when AI attempts to mimic human tone, its output retains a detectable fingerprint that distinguishes it from genuine human work.
The research, published in Nature – Humanities and Social Sciences Communications, is the first global study to apply literary stylometry to creative prose produced by AI.
Implications for Education and Publishing
While AI can produce polished, coherent text, the study cautions against using stylometry as a tool for detecting plagiarism in academic settings. Dr O’Sullivan notes that student writing varies with context and experience, making stylometric detection unreliable and ethically questionable.
For publishers and authors, the findings highlight the need for new ethical guidelines around AI‑generated content. The research underscores that the creative process—shaped by individuality and intention—remains distinct from algorithmic generation.
How UCC Is Responding
UCC’s Vice President for Research and Innovation, Professor John F. Cryan, praised the study as a testament to the university’s interdisciplinary research culture. The work aligns with UCC Futures initiatives in AI and Data Analytics, as well as the Future Humanities Institute.
Students and scholars interested in the intersection of AI and literature can explore UCC’s research portal for related projects and datasets.
Next Steps for Researchers and Practitioners
Future studies should expand datasets, experiment with new prompts, and test emerging models to refine our understanding of AI’s creative capabilities. Meanwhile, educators and publishers must balance the benefits of AI tools with the preservation of authentic human expression.
To learn more about UCC’s Digital Humanities programs and how they are shaping the future of literary research, visit the Digital Humanities course page.
Interested in collaborating with UCC researchers? Contact the research office to discuss potential partnerships.
Have questions or insights about AI writing? Write to us or share your thoughts in the comments below.
Ready to apply for a Master’s in Digital Humanities? Submit your application today.
