This systematic evidence mapping study maps and synthesises a small but growing body of empirical work exploring the crossroads between Generative Artificial Intelligence (GenAI) technologies and digital equity concerns in diverse higher education settings. A comprehensive three-stage search methodology, following systematic identification procedures aligned with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA-ScR) standards, accessed and evaluated twelve included publications from 1 January 2023 to 30 January 2025 (including online-first articles available by that date). The review extends longstanding digital divide scholarship by focusing specifically on GenAI-related configurations of access, skills, and outcomes in the post-2022 period. Additional contextualisation was provided through selective analysis of the Trust in AI Report 2025 (University of Melbourne & KPMG), which surveyed public attitudes towards AI across 47 countries and examined how employees and students use and experience AI in work and education settings, and is treated here as sector-relevant background survey evidence rather than as an additional empirical strand of the review. Analysis reveals that whilst GenAI technologies demonstrate substantial educational potential, their implementation presents considerable equity implications that warrant systematic examination. Four key barriers to equitable adoption were identified: limited digital infrastructure and connectivity constraints, insufficient artificial intelligence literacy among stakeholders, economic barriers in terms of cost affordability and access costs, and the absence of inclusive, equity-focused institutional policies. Within the limited corpus, regional disparities were emphasised as being particularly severe, with Global South case studies identifying fundamental infrastructural constraints as primary barriers to effective technology integration. Contextual survey data from the Trust in AI Report also suggest gaps between confidence in GenAI uptake and institutional readiness in many jurisdictions, particularly in emerging economies, although these findings are not specific to the higher education sector. Findings indicate that without targeted interventions to incorporate infrastructure development, comprehensive literacy programs, access provisions, and inclusive policy frameworks, GenAI deployment has the potential to exacerbate rather than alleviate existing educational inequities in the contexts studied. This review identifies the main gaps in current research and makes review-derived recommendations to policymakers, higher education institutions, and practitioners to facilitate more equitable GenAI adoption in higher education environments.

The integration of Generative Artificial Intelligence (GenAI) in academic learning has gained substantial traction across disciplines, necessitating a systematic analysis of its impact. This study explored ChatGPT's transformative role in higher education from 2022 onwards, synthesizing empirical findings across twelve distinct academic fields spanning STEM, social sciences, and healthcare. Relevant empirical case studies were identified through a systematic Scopus database search, applying discipline-specific keywords and filtering out surveys, literature reviews, and theoretical papers. Multi-stage screening identified 60 full-text articles, ultimately selecting twelve high-quality studies for rigorous cross-disciplinary analysis. The findings revealed pronounced disciplinary variations in ChatGPT adoption and impact. Quantitative analysis demonstrated that STEM disciplines report significantly higher accuracy concerns (mean = 1.57 on a 0–2 scale) compared to other fields, while healthcare disciplines showed the highest privacy concerns (mean = 2.0). Moderate positive correlation (r = 0.68) exists between academic integrity concerns and usage intensity, with computer science and social science reporting the highest levels for both metrics. Female representation, documented in 50% of studies, appears to influence adoption patterns. Sample sizes varied considerably (n = 12 to n = 430), with computer science (n = 430) and medical education (n = 265) providing robust empirical bases. Cross-disciplinary analysis revealed that ChatGPT enhances academic performance in structured problem-solving contexts, with health sciences reporting the highest positive impact scores (mean = 1.67), whilst potentially undermining critical thinking. Disciplines with text-based assessments face greater academic integrity challenges (r = 0.72 correlation).