The blurred threshold of AI-use disclosure: International journal editors’ expectations of sufficiency and necessity

Abstract

Abstract Purpose Generative AI is a powerful resource for health professions education (HPE) researchers publishing their work. However, questions remain about its use and guidance about disclosure is inconsistent. This situation is both confusing and potentially perilous for researchers, who risk their reputations if they disclose AI use inappropriately. This study explores HPE and general medical journal editors’ experiences and expectations of AI-use disclosure, in order to assist journals to clarify expectations and authors to satisfy them. Methods In this descriptive qualitative study, journal editors were interviewed between January 6, 2025, and May 7, 2025 using an online Zoom platform. Eligible participants were identified through journal webpages and snowball sampling. A purposive sampling strategy prioritized HPE research journals and included a limited sample of general medical research journals to explore transferability. Data collection and thematic analysis proceeded iteratively. Results Eighteen participants, including 9 chief editors and 9 associate/deputy editors were interviewed. Fourteen participants worked in HPE journals, four in general medical journals. The analysis revealed 4 key themes: 1) the basics of disclosure, made up of content and location expectations shared by participants; 2) the sufficiency threshold, regarding how much detail to include; 3) the necessity threshold, regarding which circumstances require disclosure; and 4) the factors blurring these two thresholds, which included the speed of change, the co-construction of disclosure standards, and the uneasy fit of scientific principles such as reproducibility and transparency with the AI-use context. Conclusions While editors shared basic disclosure expectations, they also provided insight into blurred thresholds of sufficiency and necessity that complicate disclosure. By attending to these thresholds and the factors blurring them, and by using these insights to apply recent AI-use and disclosure frameworks, journals can develop enhance their guidelines, which will assist authors in HPE in navigating the shifting norms of AI-use disclosure.

Ähnliche Arbeiten