Since its public release in November 2022, ChatGPT has been rapidly evaluated across medical disciplines, including oncology. However, the scope, methodological characteristics, and clinical focus of oncology-specific evaluations remain poorly characterized. We conducted a descriptive, cross-sectional meta-research analysis of oncology-related studies that explicitly evaluated ChatGPT, identified through Ovid Medline and Embase from database inception through December 25, 2025. Included studies were categorized by study design, oncology discipline, and artificial intelligence (AI) task, and results were summarized descriptively. A total of 1,325 oncology-related studies evaluating ChatGPT were identified, with publication volume increasing over time, including 128 (10%) in 2023, 540 (41%) in 2024, and 657 (50%) in 2025. Most studies were clinical studies (949, 72%), predominantly methodological or performance evaluation studies (658, 69% of clinical studies). Research activity was concentrated in general or multidisciplinary oncology (702, 53%) and radiation oncology (575, 43%), with limited representation in medical oncology (26, 2%) and basic science (22, 2%), and no identified studies in surgical oncology (0, 0%). ChatGPT was most frequently evaluated for diagnostic accuracy or classification tasks (760, 57%), followed by mixed AI tasks (371, 28%). Oncology-focused evaluations of ChatGPT are predominantly concentrated in methodological and performance-based study designs, with a strong emphasis on diagnostic applications and a limited range of oncology disciplines. These patterns indicate that current research has primarily focused on establishing model performance within structured evaluations, with comparatively less emphasis on broader clinical contexts. This study defines the current research landscape and provides a structured reference for interpreting how ChatGPT is being studied across oncology.
Edwards et al. (Mon,) studied this question.