LARGE LANGUAGE MODELS IN GYNECOLOGIC ONCOLOGY: APPLICATIONS AND CURRENT LIMITATIONS

Abstract

The public release of large language models has prompted rapid exploratory adoption within gynaecologic oncology, a subspecialty in which patients frequently seek accessible explanations of complex diagnoses, treatment pathways and prognosis at the point of greatest emotional vulnerability. A comparative evaluation testing two publicly available large language models against national and European gynaecologic oncology guidelines found that both models generated appropriate responses to patient-oriented questions but performed considerably less reliably on complex, guideline-dependent clinical scenarios, suggesting a role better suited to patient education than to autonomous clinical decision-making [1].