Cureus. 2026 Jun 11;18(6):e110649. doi: 10.7759/cureus.110649. eCollection 2026 Jun.
ABSTRACT
Aim Chatbots are emerging as a new and valuable tool in healthcare, offering a wide range of applications. Their use as a tool in medical physics has immense future potential. This study aimed to evaluate the performance of three artificial intelligence (AI) chatbots – ChatGPT, DeepSeek, and Gemini – in response to questions or queries related to medical physics in oncology. Materials and methods A total of 11 questions from the field of medical physics pertaining to oncology were formulated by medical physics experts. These queries were presented to the AI chatbots – ChatGPT 5.2, DeepSeek V 3.2, and Gemini 3.0 – on a predetermined date. Responses were obtained by repeating the same question once for each chatbot. The initial responses were noted and evaluated by three experts based on their correctness, completeness, ease of understanding, reliability, and applicability in the national scenario. Results The mean correctness scores were 3.4, 3.81, and 3.09 for ChatGPT, DeepSeek, and Gemini, respectively. Regarding completeness, the DeepSeek gave the maximum responses, that is, 10 complete responses to the 11 questions. No statistically significant difference was foundin real-world applicability score for the three models. Conclusion In terms of performance metrics such as correctness, DeepSeek gave better results. None of the chatbots were seen to be good enough to replicate human intelligence in metrics such as correctness, completeness, or real-world applicability. A symbiotic collaboration between AI chatbots and medical professionals is essential for enhancing healthcare delivery.
PMID:42437234 | PMC:PMC13355276 | DOI:10.7759/cureus.110649
