Feasibility of AI-Enabled Chatbots for Pre-consultation in HIV Care in Northern Nigeria.

BACKGROUND: Digital health tools are increasingly being used to support HIV care by expanding access to confidential, on-demand information. Despite growing interest, evidence on artificial intelligence (AI)-enabled chatbot use within routine HIV treatment programs in Africa remains limited. We assessed the prevalence, patterns, and determinants of AI chatbot use among people living with HIV (PLHIV) in Kano, northern Nigeria. METHOD: We conducted a cross-sectional study among 427 adults on antiretroviral treatment (ART) attending the HIV clinic in a large tertiary referral center in Kano, Nigeria. Using systematic sampling, participants were interviewed using a validated, culturally adapted interviewer-administered questionnaire. The Socio-Ecological Model and the Technology Acceptance Model guided the analyses, while multivariable logistic regression was used to identify independent predictors of HIV-related chatbot use. RESULTS: Most respondents (75.2%) were aware of AI chatbots. A similar proportion (72.6%) reported ever using one. Approximately 66.5% of participants used chatbots for HIV-related queries, most commonly to obtain general HIV/ART information (36.3%), check ART side effects (23.4%), or prepare questions for clinicians (15.0%). Independent predictors of HIV-related chatbot use included younger age (< 20 vs. ≥ 50 years: adjusted odds ratio (aOR) = 3.35; 95% confidence interval (CI), 1.12-5.21), post-secondary education (vs. none: aOR = 3.23; 95% CI, 1.58-6.25), being married (vs. divorced/widowed: aOR = 2.26; 95% CI, 1.11-6.91), shorter ART duration (< 1 year vs. > 6 years: aOR = 1.91; 95% CI, 1.15-5.67), presence of comorbidities (aOR = 2.61; 95% CI, 1.69-7.69), smartphone ownership (aOR = 2.13; 95% CI, 1.08-7.60), internet access (aOR = 6.67; 95% CI, 3.33-12.50), and English proficiency (aOR = 2.11; 95% CI, 1.13-5.45). CONCLUSION: AI-enabled chatbot use for HIV-related information is common among PLHIV receiving ART in northern Nigeria. Our findings support the need for clinician-endorsed, multilingual, and low-bandwidth chatbot designs, alongside safeguards to reduce misinformation and ensure equitable integration of AI-enabled information tools in HIV treatment programs in Africa.