Evaluating the Accuracy and Safety of AI-Driven Chatbots in Prenatal Care: A Systematic Review and Proposed Framework for Iran

The rapid adoption of artificial intelligence (AI) in maternal health has led to the proliferation of chatbots providing prenatal information. However, their accuracy and clinical safety remain uncertain. This study systematically reviewed evidence on the accuracy and reliability of AI-driven chatbots for prenatal care and proposed a validation framework for Iran. Methods: A systematic review was conducted in PubMed, Scopus, and Web of Science (۲۰۲۰-۲۰۲۵). Studies evaluating chatbot responses to prenatal queries against gold standards (clinical guidelines or expert panels) were included. Data on accuracy rates, error types, and safety mechanisms were extracted. Results: Of ۴۱۲ records, ۱۵ studies met inclusion criteria. Accuracy ranged from ۴۴% to ۹۱% across different chatbot platforms. AI-driven large language models showed lower accuracy (mean ۶۷.۳%) compared to rule-based systems (mean ۸۲.۱%). Critical errors included failure to identify emergency signs and providing outdated recommendations. Only two studies reported clinical oversight mechanisms. Conclusion: Current prenatal chatbots have inconsistent accuracy and lack standardized safety protocols. We propose a three-layer validation framework tailored to Iran's healthcare system, comprising clinical expert review, user feedback integration, and emergency triage protocols

کلیدواژه ها:

Chatbot ، Prenatal Care ، Patient Safety ، Artificial Intelligence ، Health Information Management