Abstract
Background: Chatbots and virtual health assistants (VHAs) are emerging tools in healthcare, supporting triage, symptom assessment, diagnosis, follow-up, treatment adherence, and remote monitoring. Despite the rapid adoption, comprehensive evidence on their effectiveness, limitations, and clinical utility is limited. Thus, this review evaluates current evidence regarding their applications, performance, and impact across the patient care continuum.
Methods: A systematic search of PubMed, Scopus, and Web of Science (2010–2025) identified studies on chatbots and VHAs in healthcare. Eligible studies addressed triage, symptom checking, diagnosis support, follow-up, treatment adherence, and remote monitoring. The extracted data covered design, sample size, artificial intelligence (AI) architecture, platform, usability, engagement, and outcomes. Study quality was assessed using Mixed Methods Appraisal Tool (MMAT). Given heterogeneous designs and outcomes, a narrative synthesis approach was employed to summarize the findings.
Results: Forty-six studies from 17 countries were identified, including randomized controlled trials, cross-sectional studies, pilot projects, and retrospective analyses. The reported diagnostic accuracy ranged from 33% to 93%. Usability was high (SUS 68–85), with strong potential to enhance engagement, adherence, and access to care, especially in chronic disease and mental health management. Large language model-based and retrieval-augmented systems outperformed traditional rule-based chatbots in complex tasks. Limitations included short follow-ups, simulated cases, inconsistent evaluation metrics, and scarce evidence of long-term impact.
Conclusion: Chatbots and VHAs represent effective complementary healthcare tools, improving patient support, workflow efficiency, and accessibility, although they should not be independently used in diagnostic/therapeutic decision-making. Future research should emphasize large-scale real-world evaluations, standardized metrics, long-term outcomes, and secure, patient-centered implementations.