Summary
Provides consensus best practices and evaluation methods for chatbots that provide reliable nonclinical health information, personalized guidance, health-literacy support, and triage or escalation without making clinical recommendations.
Healthcare Implications
Developers and implementers should define nonclinical scope, test answer quality and usability, disclose limitations, validate triage and escalation behavior, assess safety and subgroup performance, monitor failures and drift, and prevent the chatbot from presenting itself as a substitute for professional care.