(068) Can AI Chatbots Care? Evaluating ChatGPT and Gemini in Addressing the Mental Health Aspects of Female Sexual Dysfunction

Abstract Introduction Artificial intelligence (AI) chatbots such as ChatGPT and Gemini are increasingly used by women seeking reliable information and emotional support regarding sexual health. Beyond physical symptoms, many women face psychological challenges including anxiety, guilt, and low self-esteem. As these platforms become more integrated into patient information-seeking behaviors, it is essential to evaluate their capacity to address the psychological and emotional dimensions of female sexual dysfunction with accuracy, empathy, and safety. Objective This study aimed to evaluate and compare ChatGPT and Gemini in addressing the mental health aspects of female sexual dysfunction, focusing on information accuracy, empathy, recognition of emotional distress, safety, and the quality of next-step guidance. Methods Twenty-five standardized questions related to female sexual dysfunction were presented to ChatGPT (version 5 omni ChatGPT-5o, OpenAI) and Gemini (Google, free version). The questions covered five major domains: desire, arousal, orgasm, pain, and emotional impact. Three urologists independently evaluated all responses. Information quality was assessed using the DISCERN instrument (1 = low, 5 = high). Understandability and actionability were measured with the Patient Education Materials Assessment Tool for Printable Materials (PEMAT-P; 0% = low, 100% = high). Misinformation was rated from 1 (no misinformation) to 5 (high misinformation). Additional parameters included Safety (1–5), Empathy and Tone (0–2), Emotional Distress Identification (0–2), and Quality of Guidance (1–5). Statistical and graphical analyses were performed using RStudio (version 2023.12.1 + 402). Results Mean DISCERN scores were similar between ChatGPT and Gemini (4.78 vs 4.76, p = 0.26). PEMAT–Understandability was significantly higher for Gemini (99.0% vs 96.1%, p = 0.013), whereas PEMAT–Actionability showed no significant difference (89.1% vs 93.6%, p = 0.71). Misinformation scores were comparable between models (1.24 vs 1.22, p = 0.86), and safety ratings remained high for both (4.66 vs 4.62, p = 0.52). ChatGPT demonstrated greater empathy (1.64 vs 1.32, p = 0.011), more frequent identification of emotional distress (1.64 vs 0.96, p 0.001), and higher quality of guidance scores (4.68 vs 4.34, p = 0.008). Conclusions Both ChatGPT and Gemini provided high-quality, safe, and reliable responses with minimal misinformation. Gemini demonstrated superior understandability, while ChatGPT excelled in empathy, recognition of emotional distress, and quality of guidance. These findings indicate that although both models are suitable for patient education, ChatGPT may offer greater value in addressing emotional and psychosocial aspects of care for women with sexual dysfunction. Disclosure No.