Abstract Background The integration of generative artificial intelligence (GenAI) tools into higher education is rapidly accelerating, yet the educational ecosystem appears largely unprepared for the scale and speed of adoption. Our recent international survey of over 1,200 university students and postdoctoral fellows in the medical and life sciences found that while more than 80% had used GenAI tools, the vast majority had received no formal training, and many were unaware of any institutional policies governing their use. This gap between adoption and institutional preparedness may pose risks to research integrity if left unaddressed. Main body Existing research supports a coordinated, evidence-informed approach to AI education in research performing organizations (RPOs), grounded in principles of research integrity, equity, and critical engagement with technology. First, GenAI literacy must be embedded as a core competency in scientific and clinical education, equipping students to critically evaluate tool outputs, recognize hallucination and bias, and understand the distinction between AI-generated inaccuracy and deliberate research misconduct. Second, traditional frameworks for academic integrity would benefit from re-examination. Today, GenAI tools have fundamentally disrupted established notions of authorship, originality, and plagiarism, and the absence of coherent institutional guidance leaves students and educators navigating these questions without support. Third, RPOs must develop clear, transparent, and context-sensitive policies on GenAI use by embedding responsible AI education into curricula, equipping learners not only to use these tools effectively, but to question, verify, and critique their outputs in ways that uphold scientific standards. Conclusion RPOs have an urgent responsibility to move beyond reactive responses and toward proactive strategies that uphold scientific standards, promote equity of access, and prepare the next generation of researchers and clinicians to engage with GenAI responsibly and ethically.
Large language models (LLMs) are increasingly used across scholarly writing, peer review, and editorial workflows, creating new research integrity questions about attribution, accountability, and disclosure. We measured how explicit artificial intelligence (AI) transparency statements are appearing in published medical literature and whether scalable full-text methods can monitor their adoption across journals and specialties. We conducted a retrospective cross-sectional bibliometric analysis of 97,916 English-language medical journal articles published from 2015 through April 2026 across 14 medical and surgical fields. Full text was retrieved from PubMed Central XML or publisher-hosted article files and classified with a large language model for positive AI-use disclosure, explicit statements of non-use, named tools, use cases, and statement location. Validation used pre-ChatGPT specificity, second-model agreement in 10,979 articles, and blinded human review of 300 stratified cases by 2 independent reviewers with adjudication. All rates were recalculated excluding Frontiers journals, whose submission workflow applies a standardized AI declaration. Any AI transparency statement increased from 1.4
Linguistic inequity affecting non-native English speakers is widely documented in academia. This inequity is especially visible in philosophy, where success depends unusually heavily on the quality and nuance of written expression. Yet existing studies rely on small samples from a handful of elite journals, limiting their generalizability. We address this gap with the most comprehensive empirical analysis to date, covering 103,595 articles published across 223 journals between 1975 and 2021. Using two automated classifiers based on institutional affiliation and PhD country (accuracy above 89
Prospective study registration is a key approach to mitigate reporting bias. Observational studies are not routinely registered. We aimed to estimate the prevalence of prospective registration among observational studies of health care interventions, explore authors’ attitudes toward registration, examine factors associated with prospective registration, and chart editorial policies on registration of observational studies in medical journals. We conducted a cross-sectional meta-research study comprising three components: 1) We sampled 200 PubMed-indexed cohort or case–control study publications assessing the effects of health care interventions, published in 2023, searched for corresponding publicly availably registration records, and surveyed the study authors on attitudes to registration; 2) We evaluated associations between prospective registration (≤ 1 month of study start) and study characteristics by multivariable logistic regression; 3) We examined editorial policies of 40 medical journals to identify requirements to register observational studies, and invited one editor per journal to share their perspectives on registration. We included 200 publications: 197 cohort studies and 3 case–control studies. In total, 17 (9
Current research evaluation metrics emphasize publication quantity, creating incentives for rapid outputs. Letters to the editor (LTEs), a usually brief form of post-publication peer review, are increasingly exploited under the pressures of the “publish or perish” culture because they are citable, lightly regulated, and often bypass the peer-review process. Their textual format, typically short, minimally referenced, and frequently not peer reviewed, makes LTEs particularly easy to generate using large language models, raising concerns about the potential for artificial intelligence (AI)-assisted publication inflation. Since the introduction of AI chatbots in 2022, LTE submissions have surged, with some authors producing rapidly increasing numbers of letters. We investigate the role of generative AI in LTE production and propose strategies to mitigate this emerging issue. We downloaded more than 780,000 LTE records indexed in PubMed from 2005 to 2025. Using descriptive statistics, we analyzed temporal trends and interpreted them using an epidemiological perspective, classifying authors exhibiting unusually rapid growth in correspondence output, and assessed their proportional contributions to annual LTE publications. The mean annual number of LTEs per author shows three distinct accelerated growth years: in 2013, associated with MEDLINE indexing changes; in 2020, coinciding with the COVID-19 pandemic; and in 2023 to date, aligning with the widespread adoption of generative AI. From an epidemiological perspective, the post-LLM period is marked by an increase in the prevalence of high-output authors, with newly debuting researchers more likely to produce multiple LTEs in their first year. The study identifies 2023–2025 as a period of anomalous LTE growth associated with the rise of AI language models, suggesting that AI may contribute to publication inflation and highlighting the need for safeguards in scientific communication.
