Key points are not available for this paper at this time.
The integration of AI technologies in academic writing has become an increasingly prevalent practice. However, existing discussions have primarily focused on analyzing practitioners’ perceptions of AI use in academic writing, with limited attention given to authentic examples of AI-generated content in such practices. This study analyzes 91 instances of undeclared AI-generated content in published research articles listed on Retraction Watch, a watchdog organization that tracks retractions of scientific papers. The results indicate that undeclared uses of AI serve three primary functions: adding new content (49%), organizing existing content (33%), and performing language and formality checks (18%). These AI-generated instances are most frequently found in introduction and literature review sections (34%), followed by results and discussion sections (31%), and least in methodology sections (22%) of research articles. The study raises ethical concerns about the undeclared use of AI, underscoring its potential to compromise academic integrity and knowledge production within the academic community.
Yao et al. (Thu,) studied this question.