Research Spotlight No. 15
Can ChatGPT Write Like Researchers? A Comparative Study of Medical Research Article Titles
Two researchers from the Department of English Language and Literature at the Faculty of Arts, Beni Suef University, conducted a study on the ability of generative artificial intelligence, represented by ChatGPT, to simulate researchers in formulating scientific research titles. This was done through comparing medical research article titles written by human researchers with corresponding titles generated by ChatGPT.
The study was based on two parallel corpora comprising 600 research titles. The first corpus consisted of 300 article titles written by researchers and published in high-impact medical journals: The Lancet, The BMJ, and JAMA. The second corpus consisted of 300 titles generated by ChatGPT based on the abstracts of the same published research articles.
The study compared the two groups in terms of title length, structure (single-unit or double-unit titles), linguistic composition, and semantic focus.
The findings revealed a strong similarity between human-authored and ChatGPT-generated titles in average title length and in adherence to the nominal structure commonly used in medical discourse. There was also a clear presence of multi-unit titles separated by colons, reflecting ChatGPT’s ability to absorb prevailing conventions in academic medical writing.
However, the analyses also showed notable differences. ChatGPT tended to use double-unit titles more frequently and placed greater emphasis on methodology, data, and results. In contrast, human researchers demonstrated clearer stylistic diversity and rhetorical flexibility, particularly in the use of concise topical titles or audience-oriented titles.
These findings indicate that ChatGPT has a high ability to imitate prevailing academic writing patterns, yet it also tends toward formulaic and standardized styles. If relied upon uncritically, this may reduce rhetorical diversity in scientific discourse.
The study concludes that artificial intelligence can be an effective tool for supporting academic writing, especially in adhering to formal conventions. However, responsible use requires conscious human supervision that balances compliance with scientific standards while preserving the uniqueness, style, and research vision of each study.
The research was published in Humanities and Social Sciences Communications, affiliated with Springer Nature. The journal has an impact factor of 3.6, is ranked in Q1, and is among the top 10% in the field of humanities according to Web of Science.
English Brief:
This article examines the extent to which generative artificial intelligence, specifically ChatGPT, can replicate human practices in academic writing by focusing on the micro-genre of medical research article titles. Using two parallel corpora comprising 300 human-authored titles from high-impact medical journals and 300 ChatGPT-generated titles based on the same abstracts, the study compares title length, structural form, syntactic patterns, and content focus. Quantitative and qualitative analyses reveal a strong convergence between human and AI-generated titles, particularly in title length, preference for multi-unit structures, and dominance of nominal constructions. Both corpora largely favor methods-focused titles, reflecting disciplinary norms in medical research. However, ChatGPT shows a stronger tendency toward formulaic, multi-unit titles and places greater emphasis on datasets and results, while human authors display more rhetorical flexibility and stylistic variation. The findings suggest that ChatGPT can accurately model established genre conventions but may also reinforce standardization, underscoring the need for critical, informed human oversight when using generative AI in academic writing.
Research Title and Link:
Generative AI in academic writing: a comparison of human-authored and ChatGPT-generated research article titles
[https://doi.org/10.1057/s41599-026-06956-z](https://doi.org/10.1057/s41599-026-06956-z)