The latest article in Nature reveals the extent of the hidden use of ChatGPT in academic writing. Investigators detected characteristic chatbot phrases in hundreds of publications — inserted without disclosure — undermining the credibility of the peer-review process. The Nature article cites, among others, research conducted by our professor, who was the first to document such cases in top-quartile Scopus-ranked journals.
The Nature editorial highlights three key phenomena:
-
Researchers identified characteristic phrases such as "As an AI language model" or "Regenerate response" in hundreds of papers across various fields, revealing the involvement of ChatGPT.
-
The risk of an error cascade — undisclosed AI-generated passages are already being cited in the literature, potentially multiplying unverified information.
-
Nature contacted publishers including Springer Nature, Taylor & Francis, IEEE, and others named by the researchers. Editors reported that all flagged articles are currently under review.
The first comprehensive analysis, published in December 2024 and authored by a professor from the University of Economics in Katowice, demonstrated that undisclosed AI content appears even in top-cited journals — and some of these articles have already been cited elsewhere, highlighting the systemic nature of the problem. Nature also reports on the phenomenon of "silent corrections" — where publishers quietly remove chatbot-generated phrases from already published articles without issuing official errata.
A discussion is ongoing about when and how authors and reviewers should disclose the use of AI tools. Although publisher policies vary, they all agree on one principle: full transparency is essential to safeguard trust in scientific literature.
Research conducted by Prof. Artur Strzelecki from the Department of Informatics — and widely discussed by Nature — shows that the University of Economics in Katowice remains at the forefront of the global debate on AI transparency in science.
The full article, titled “Science sleuths flag hundreds of papers that use AI without disclosing it”, is available at www.nature.com/articles/d41586-025-01180-2 (DOI: 10.1038/d41586-025-01180-2), accessible through the UE Library.
Dołącz do nas