Artificial intelligence adoption in higher education institutions of low-income countries

Abstract Debates on artificial intelligence (AI) in higher education remain dominated by evidence from high-income contexts, often assuming institutional capacity and robust infrastructure. This literature review redirects attention to low-income countries (LICs), where structural constraints profoundly condition the opportunities and risks of AI adoption. Guided by the PRISMA methodology, a systematic search was conducted for the period 2022–2024. The resulting body of scholarship was analysed through the lens of Rogers’ Diffusion of Innovations framework to examine patterns of adoption, enabling and constraining factors, and trajectories of diffusion. The evidence revealed early but fragile patterns of adoption. AI demonstrated its greatest relative advantage in targeted applications, such as chatbots for vaccine literacy in higher education institutions, AI-assisted examinations, and generative AI tools, which yield immediate, visible benefits in terms of learning outcomes, motivation, or efficiency. Yet systemic integration remained elusive. Compatibility with local pedagogical traditions was limited, infrastructural deficits and weak digital capacities heightened perceived complexity, and institutional governance lagged behind technological uptake. Opportunities for trialability and observability emerge as decisive enablers: where low-cost pilots provided tangible evidence of impact, adoption gained traction. The review advances two arguments: first, in LICs the drivers of AI adoption are pragmatic, problem-focused, and contingent rather than transformative; second, without deliberate investment in infrastructure, faculty development, and context-sensitive governance, AI risks entrenching dependency rather than fostering equity.