AI-Enabled Chatbot Interventions on Health Outcomes among People Living with HIV: A framework-guided Systematic Review

Background Artificial intelligence (AI)–enabled chatbots are increasingly proposed as scalable tools to support linkage to HIV care, ART treatment adherence, and mental health among people living with HIV (PLWH). However, it remains unclear whether existing chatbot interventions are sufficiently developed, evaluated, or ethically governed to meaningfully improve outcomes for PLWH. Prior reviews have examined digital or mobile health tools broadly, but limited efforts have systematically assessed chatbot interventions through AI-specific implementation and governance frameworks. Methods We conducted a systematic review (PROSPERO: CRD420251271843) following the PRISMA 2020 guidelines across eight databases, covering publications from January 2005 to December 2025. Eligible studies examined the development, implementation, or evaluation of chatbot interventions designed to support health outcomes among PLWH. Data extraction and synthesis were guided by implementation and AI-specific frameworks, including SPIRIT-AI, CONSORT-AI, TEHAI, and WHO guidance on AI ethics and governance. Results Ten studies published between 2020 and 2025 met the inclusion criteria, representing 138 participants across diverse populations, including PLWH (adolescents and adults), caregivers, and healthcare providers, primarily from North and South America. Chatbots were designed to assist HIV management through ART adherence support, appointment reminders, resilience building, peer support promotion, healthcare provider access and connection, disclosure decision-making, and psychoeducation, with the majority being mobile- or web-based and using natural language processing or rule-based dialogue systems, with limited use of large language models. While usability, acceptability, and feasibility outcomes were consistently favorable, rigorous evaluation of clinical or mental health outcomes was largely absent. Framework-guided assessment revealed substantial gaps in reporting on potential harms, real-world integration, and adoption readiness, indicating limited alignment with established AI implementation and governance standards. Conclusions To the best of our knowledge, this is the first systematic review of AI-enabled chatbot interventions for PLWH, which highlights a critical gap between technological innovation and clinical impact. Despite growing enthusiasm for AI-enabled chatbots in HIV care, the current evidence base remains largely developmental and insufficient to support scale-up or policy adoption. Future research must move beyond usability testing toward ethically grounded, framework-aligned evaluations to translate promising digital innovations into scalable, ethical, and sustainable tools that can advance long-term HIV treatment outcomes.