Artificial Intelligence and Anxiety Care: Clinical Promise and Ethical Boundaries
Artificial intelligence is increasingly used in mental health screening, symptom monitoring, conversational support, and low-intensity digital interventions. This structured critical narrative review examines the clinical promise and ethical boundaries of artificial intelligence in anxiety care. A targeted search of PubMed/MEDLINE, official publisher platforms, institutional guidance, and reference lists was updated through June25,2026. Sources were assessed across seven analytical domains: technological class and intended function, anxiety-related applicability, strength of effectiveness and safety evidence, user experience and persuasive design, ethical and professional risk, equity, and human oversight. Evidence indicates that selected conversational interventions can produce small-to-moderate short-term reductions in anxiety symptoms, but findings are heterogeneous and cannot be generalized from bounded, clinically curated systems to unrestricted consumer chatbots. Persistent risks include delayed escalation, automation of empathy, reassurance dependence, privacy concerns, subgroup bias, and unequal digital access. Artificial intelligence is therefore most defensible as a task-specific, clinically delimited support within stepped or blended care, with transparent role definition, version-specific validation, continuous safety monitoring, and accountable human oversight.