Perceptions Of Generative AI in the Global South: A Scoping Review

Synthesis: Nguyen and Perkins (2026) conduct a scoping review of 75 papers (2022–2025) synthesizing current perceptions of Generative AI in higher education across the Global South. Following the PRISMA-ScR methodology, they categorize findings into five areas: GenAI acceptance and adoption, implications and challenges, academic integrity considerations, educational practices, and equity concerns. GenAI offers transformative possibilities for personalized learning, research support, and administrative efficiency, yet its implementation is hampered by infrastructure limitations, human capital deficiencies, ethical concerns, inadequate policy frameworks, and contextual challenges. Notably, equity has received the least research attention despite its critical importance to inclusive education. The review identifies substantial gaps — limited geographic representation, stakeholder imbalance, and insufficient exploration of long-term outcomes — and urges equity-centered, context-specific, interdisciplinary research.

Key Findings

Five research areas. The 75 included studies cluster into five principal categories: (1) GenAI acceptance and adoption; (2) implications, possibilities, and challenges; (3) academic integrity and ethical considerations; (4) GenAI in educational practice and competency development; and (5) equity concerns.

Acceptance and adoption. Nineteen papers examine adoption drivers, largely through technology-acceptance models such as UTAUT, TAM, and Diffusion of Innovation. Performance expectancy is the most consistent predictor of adoption, while the influence of effort expectancy, social influence, trust, perceived risk, and perceived ease of use varies markedly by context (e.g., Philippines, India, Peru, China, Indonesia, Thailand, UAE, Nigeria). Adoption is also shaped by gender, academic discipline, geographic location, cultural dimensions, and cognitive readiness.

Implications and challenges. GenAI is perceived to support research assistance, personalized feedback, assessment support, curriculum design, and academic literacy, but also to introduce risks of plagiarism, misinformation, over-reliance, bias, and erosion of deep learning. Implementation faces cross-cutting barriers: infrastructure and resource limitations, human capital and expertise deficiencies, inadequate policy frameworks, data-privacy concerns, and cultural or linguistic challenges. Governance approaches vary sharply across national contexts.

Academic integrity and ethics. Studies reveal a gap between students' stated ethical positions and their actual AI-usage behaviors — AI-assisted cheating may be nearly three times higher than direct self-reports suggest. Students strongly disapprove of directly copying AI output yet are uncertain about subtler forms of assistance, prompting the concept of "AI-giarism." Existing ethical frameworks are critiqued for addressing only surface-level problems, with calls to decolonize AI ethics (e.g., via Ubuntu philosophy) and to build educator ethical competencies.

Educational practice. GenAI supports personalization, self-regulation, competency development, assessment reform (e.g., the Artificial Intelligence Assessment Scale), and English language education, but tensions persist around over-reliance, reduced cognitive independence, and the need for critical engagement with AI outputs. Educators need new competencies and institutional support.

Equity — the least-studied area. Digital-divide and access disparities, gender and intersectional biases, policy/governance, and epistemic justice receive the least research attention despite their centrality to inclusive education. The review recommends auditing digital access, subsidising GenAI for underserved students, and tracking impacts on different learner groups.

Study Design & Method

The study is a scoping review aligned to the PRISMA-ScR checklist (Tricco et al., 2018), designed to map concepts and gaps rather than evaluate effects. Searches ran across SCOPUS, ERIC, Web of Science, and Google Scholar (supplementary) using a Boolean string covering the technology, educational context, stakeholder perceptions, and geographic focus, for the period 12/2022–2/2025 in English. The Global South is defined via UNCTAD's (2018) classification of developing economies. Screening narrowed 777 records to 102 full texts, with 75 studies included. Analysis used inductive category development from abstracts followed by deductive full-text refinement, assisted by a GenAI tool (Claude Pro) under the ACTOR framework, with all AI output reviewed and verified by human researchers; a light appraisal excluded methodologically weak studies, and no formal risk-of-bias assessment was undertaken.

What this means for practice

• Administrators. Write GenAI policy that reflects local infrastructure, languages, and cultures rather than importing Western defaults, and pair it with sustained professional development in AI, Pedagogies and Teaching Strategies, and Ethics; the review finds governance approaches vary sharply across national contexts.
• Administrators. Audit and subsidize access before scaling deployment: track how GenAI reaches different learner groups and support underserved students, since equity is the least-studied of the review's five areas across 75 included studies.
• Instructors. Teach AI Literacy and Academic Integrity explicitly and redesign assessment so students critique, verify, and justify AI outputs, given the gap between students' stated ethical positions and their reported behavior and the murkier territory of "AI-giarism."
• Designers. Align GenAI-supported tasks with assessment-reform frameworks such as the Artificial Intelligence Assessment Scale, pairing any AI assistance with activities that require students to evaluate what the tools produced.
• Researchers. Prioritize equity-centered, context-specific, methodologically diverse work and stop treating the Global South as a single category; the included literature is dominated by student perspectives and by Western-developed acceptance models such as UTAUT and TAM.

Limitations

• Screening was not duplicated: a single reviewer screened all 777 records, which the authors flag as increasing the risk of missed eligible studies (102 full texts assessed, 75 included).
• No formal critical appraisal or risk-of-bias assessment was performed because the aim was to map a heterogeneous evidence base spanning peer-reviewed articles, conference papers, preprints, and gray literature; only a light appraisal removed the weakest studies.
• The search covered four databases for 12/2022-2/2025 in English only, and the design maps concepts and gaps rather than evaluating effects, so no claim about GenAI's impact on learning outcomes can be drawn from it.
• The evidence base is skewed toward student perspectives and flattens very different educational systems into one category, limiting applicability across regional and national settings.

Citation

Nguyen, A. T., & Perkins, M. (2026). Perceptions of Generative AI in the Global South: A Scoping Review. Journal of University Teaching and Learning Practice, Advanced Online Publication. https://doi.org/10.53761/f22j6648 (CC BY-ND 4.0)