Artificial intelligence in pain: A comprehensive review

Pain is a major public health problem worldwide due to its high prevalence and substantial negative impact on quality of life. Artificial intelligence (AI)-based chatbots are increasingly used to access health-related information; however, evidence regarding the readability, quality, reliability, and alignment of pain-related information provided by these systems with clinical practice guidelines remains limited and heterogeneous. This review aimed to comprehensively examine the existing literature evaluating responses generated by AI-based chatbots in the field of pain with respect to readability, information quality, reliability, and adherence to clinical practice guidelines. Studies published between 2024 and 2025 that assessed AI chatbot responses to pain-related questions were analyzed. ChatGPT, Gemini, Perplexity, DeepSeek, and other large language models were evaluated. Readability was assessed using the Flesch Reading Ease, Flesch-Kincaid Grade Level, and SMOG indices, while information quality and reliability were evaluated using DISCERN, the JAMA Benchmark Criteria, EQIP, and the Global Quality Score. Adherence to clinical guidelines was examined through comparisons with relevant national and international recommendations. The reviewed studies demonstrated that AI models generally provide accurate basic information about pain. However, most responses exceeded the recommended readability levels for patient education materials. Information quality and reliability were typically rated as moderate, with reported deficiencies in the discussion of treatment risks, alternative options, and source transparency. Although adherence to clinical guidelines was acceptable at the level of general principles, inconsistencies were identified in diagnostic details and treatment sequencing. While AI-based chatbots show potential as supportive tools for pain-related information, their current use as independent sources for clinical decision-making or primary patient education is limited. Human oversight and the generation of health literacy-appropriate content are essential for the safe use of these systems.