AI-Based Conversational Agents Show Potential to Improve Hypertension and Type 2 Diabetes Management: A Systematic Review (Preprint)

<sec> <title>BACKGROUND</title> <p>Artificial intelligence (AI)–based conversational agents (CAs), including chatbots and virtual assistants, are increasingly being used to support chronic disease self-management. However, evidence regarding their implementation, effectiveness, and equity implications in management of hypertension and type 2 diabetes has not been synthesized.</p> </sec> <sec> <title>OBJECTIVE</title> <p>To synthesize evidence on the feasibility, acceptability, implementation, effectiveness, and equity of AI-based conversational agents for adults with hypertension and/or type 2 diabetes.</p> </sec> <sec> <title>METHODS</title> <p>We conducted a systematic review in accordance with PRISMA 2020 guidelines. EMBASE, Web of Science, MEDLINE, and CINAHL were searched for studies published between 2015 and 2025. Google Scholar and reference lists of relevant reviews and included studies were searched to identify additional records. Eligible studies evaluated AI-based conversational agents among adults with hypertension and/or type 2 diabetes and reported clinical, behavioral, patient-centered, or implementation-related outcomes. Study quality was assessed using the Mixed Methods Appraisal Tool, and findings were synthesized narratively following the Synthesis Without Meta-analysis guidelines. The certainty of evidence was assessed using the GRADE approach.</p> </sec> <sec> <title>RESULTS</title> <p>Sixteen studies met the inclusion criteria, encompassing diverse study designs and settings across 4 low-/middle-income and 11 high-income countries. Sample sizes ranged from 10 to 64,679 participants, primarily including adults with type 2 diabetes, hypertension, obesity, or cardiometabolic conditions. AI-based conversational agents were delivered through mobile applications, SMS/text messaging, and other messaging platforms. Interventions commonly supported medication adherence, blood pressure and glucose monitoring, lifestyle modification, psychosocial support, and self-management education. Statistically significant improvements were reported in glycemic outcomes, with HbA1c reductions ranging from 0.30% to 1.04%. Blood pressure benefits were observed primarily for systolic blood pressure and overall blood pressure control, with systolic blood pressure reductions ranging from 3.0 to 6.5 mmHg. Behavioral and patient-centered outcomes also improved, including medication adherence (60.0% to 82.2%), depression symptoms, diabetes distress, and health-related quality of life. Implementation findings indicated high feasibility and acceptability, particularly when conversational agents were integrated into structured care pathways involving nurse- or pharmacist-led support. Equity-related findings indicated that CAs were implemented among low-income, uninsured, older, and ethnically diverse populations, but digital access and literacy remained barriers; equity-promoting strategies included SMS delivery, multilingual content, and subsidized devices. Declining engagement over time was also reported. Overall, certainty of evidence ranged from very low to low due to methodological limitations, heterogeneity, incomplete outcome data, and potential publication bias.</p> </sec> <sec> <title>CONCLUSIONS</title> <p>AI-based conversational agents show promise for improving self-management, behavioral outcomes, and selected clinical outcomes among adults with hypertension and type 2 diabetes. Their effectiveness may be greater when integrated into broader healthcare systems and sustained models of care. Addressing digital access, literacy, and representation will be essential to ensure equitable implementation and benefit. More rigorous, large-scale studies with longer follow-up are needed to evaluate clinical effectiveness, sustainability, implementation, and equitable access across diverse populations.</p> </sec>