Endometriosis information generated by ChatGPT and Gemini: expert-rated accuracy and communicative adequacy

Dataset from a cross-sectional comparative evaluation study of large language model responses about endometriosis, rated blindly by specialists from the Brazilian Endometriosis Society. Twenty questions — ten framed from the perspective of patients and ten from the perspective of gynecologists without specific experience in the disease — were submitted to ChatGPT (GPT-4o) and Gemini (1.5 Pro) in June 2024, in Brazilian Portuguese. The 40 responses were segmented into 190 thematic parts and evaluated for technical accuracy and communicative adequacy by 16 expert raters between October 2024 and February 2025. Approved by the Research Ethics Committee of the Federal University of São Paulo (protocol 6.657.855).