Can ChatGPT be trusted? Evaluating AI responses to oral health questions among pregnant Arabic-speaking women.

BACKGROUND: ChatGPT, an artificial intelligence (AI) chatbot developed by OpenAI, is increasingly being used in healthcare, including dentistry, for patient education; this study aimed to assess the usability and quality of ChatGPT's responses to pregnancy-related oral health queries in Saudi Arabia. METHOD: This two-part cross-sectional study assessed pregnant Arabic women's perceptions of ChatGPT for oral health queries and evaluated its responses using an online questionnaire. Responses from ChatGPT-4o mini were rated by 5 dental experts with regard to accuracy, clarity, relevance, and acceptance using a 5-point Likert scale. RESULTS: Among the 300 participants, 42.0% (126) knew about ChatGPT, 33.7% (101) had previously used it, 14.3% (43) had used it to obtain medical information, 8.7% (26) had used it for dental information, and 8.3% (25) had used it for dental information during pregnancy. Attitudes regarding ChatGPT were rated from 1 to 4. Except for 1 item, the means were all above the midpoint. Attitude ratings ranged from a mean of 2.71 (SD 0.76) for ChatGPT competency to a mean of 2.34 (SD 0.92) for its ability to replace human interactions. However, ChatGPT competency (P = .028), security (P = .015), willingness to use ChatGPT for inquiries (P = .021), ability to assist in informed decision-making (P = .01), willingness to make decisions based on recommendations (P = .024), and persuasiveness (P = .049) were significantly different based on educational level. Pregnant women with higher levels of education rated these aspects significantly lower than those with a high school diploma or bachelor's degree. CONCLUSION: ChatGPT provided useful oral health information for pregnant individuals, but its responses required revision and supervision by health professionals. Its usage among pregnant women in Saudi Arabia remained low.