Reliability of Artificial Intelligence Platforms as Patient Information Source in Orthodontics- A Cross-sectional Study

Introduction: Artificial intelligence (AI)–based chatbots are increasingly accessed by patients for orthodontic information. Several studies have evaluated the technical performance of these tools. Evidence comparing patient perception with expert assessment remains limited. This study aimed to assess and compare the accuracy, reliability, and quality of orthodontic information generated by three AI chatbots. Methods: A cross-sectional study was conducted among 156 orthodontic treatment–seeking patients at a tertiary care center in Nepal. Three frequently asked orthodontic questions (FAQs) related to braces pain, clear aligners, and mini-implants were presented to ChatGPT, Google Gemini, and DeepSeek by the primary investigator. Each chatbot’s responses to the FAQs were evaluated by the patient for its accuracy using a five-point Likert scale, for its reliability using a Modified DISCERN tool, and for its overall Quality using the Global Quality Scale (GQS). An expert orthodontist independently assessed and scored all the responses of the chatbot as standard reference. Statistical analysis was performed using SPSS version 22, with significance set at p<0.05. Results: Patient-rated scores showed statistically significant differences from expert scores for most responses across all three domains (p<0.05). ChatGPT consistently received lower scores for overall accuracy, reliability, and quality. Google Gemini and DeepSeek demonstrated significantly higher patient-rated scores, with no significant differences between them. Reliability scores were comparatively lower across all platforms, particularly when assessed using the Modified DISCERN tool. Conclusions: AI chatbots provide basic orthodontic information that patients find understandable; however, their accuracy, reliability, and clinical completeness remain inferior to expert standards.