ChatGPT-based conversational AI tool for training oncologists in patient communication: A real-world implementation study.

Abstract

e13675 Background: Oncologists frequently experience difficulties communicating with patients, particularly when delivering bad news and discussing prognosis. Effective communication training helps physicians conduct these conversations more efficiently and with greater confidence. We developed an artificial intelligence-powered chatbot to support oncologists in patient communication. Methods: We created a chatbot based on the large language model ChatGPT 4.1 integrated into Telegram messenger (@MedcomAI_bot). The bot contains contextual information on established models and protocols of medical communication, including the Calgary-Cambridge consultation guide, the SPIKES protocol for delivering bad news, the CONES protocol for communicating medical errors, and the NURSE acronym for demonstrating empathy. The chatbot also incorporates assessment checklists to evaluate consultations according to these frameworks and evidence-based research in medical communication.The bot operates in three modes: (1) assisting physicians in planning conversations with patients, (2) providing feedback on completed patient interactions, and (3) assessing conversations based on provided transcripts. Results: Between October 2025 and December 2025, 301 physicians utilized the @MedcomAI_bot chatbot, generating 417 distinct dialogues. Of these interactions, 105 (25.2%) received positive user ratings, 306 (73.4%) were neutral, and 6 (1.4%) received negative ratings. The tool demonstrated high engagement, with an average of 1.39 dialogues per physician. Preliminary analysis suggests particular utility in preparing for difficult conversations and obtaining real-time feedback on communication performance. Conclusions: AI-powered conversational tools show promise as scalable, accessible educational resources for training oncologists in medical communication and patient counseling. Further prospective studies are needed to assess impact on communication competency and patient satisfaction outcomes.