Generative artificial intelligence in higher education: A systematic review of perceptions, implementation and pedagogical transformation

Abstract The integration of generative artificial intelligence (GAI) in higher education presents significant opportunities and challenges, yet a systematic understanding of its multifaceted impact remains fragmented. This systematic review synthesizes empirical evidence on GAI implementation to propose an integrated framework of best practices. We address four research questions examining student and faculty perceptions, institutional integration strategies and barriers, ethical risks and pedagogical innovation potential. Following PRISMA methodology and guided by the TPACK framework, we analysed 125 empirical studies published between 2021 and 2025. Our findings reveal four distinctive patterns. First, the ‘curriculum integration paradox’ shows that institutional investments in faculty development yield a negligible correlation ( r = 0.12) with pedagogical transformation. Second, geographical analysis uncovers divergence between the Global North's focus on creativity enhancement and the Global South's challenges with fundamental access. Third, ethical risks form an interconnected ecosystem where academic integrity, privacy, bias and equity interact complexly. Fourth, persistent tension between enthusiasm (73% student adoption) and faculty resistance reflects deeper epistemological uncertainties. Discussion highlights that GAI's transformative capacity extends beyond automation to redefining pedagogical roles and fostering metacognition, yet realizing this potential requires comprehensive ecosystem development. We conclude that successful GAI integration demands more than technological adoption—it requires fundamental transformation of educational paradigms through contextually sensitive, ethically grounded strategies that balance innovation with equity. The proposed framework emphasizes systemic integration and participatory governance to create inclusive, effective educational futures. Context and implications Rationale for this study: The rapid integration of generative artificial intelligence (GAI) in higher education lacks an updated synthesis addressing systemic patterns, geographical disparities and the gap between institutional investment and actual pedagogical transformation. Why the new findings matter: Our findings reveal a ‘curriculum integration paradox’ and an interconnected ethical ecosystem, providing an empirical framework to guide evidence‐based policies and practices in GAI implementation. Implications for practitioners, policy‐makers, and researchers: For practitioners, findings emphasize evolving from knowledge transmitters to learning experience designers, leveraging GAI to foster critical thinking and metacognition. For policymakers, results underscore the urgency of developing ethical, inclusive and context‐sensitive governance frameworks that address global digital divides while ensuring educational equity. For researchers, we identify critical gaps, including the absence of longitudinal studies and the need for integrated theoretical models connecting technology, pedagogy and ethics within expanded TPACK frameworks. The curriculum integration paradox particularly demands intervention studies exploring how professional development can translate into meaningful pedagogical transformation across diverse educational contexts.