Generative Artificial Intelligence in Higher Education: A Systematic Review of Tools, Learning Outcomes, and Factors Influencing Integration
Abstract
Generative Artificial Intelligence (GenAI) is reshaping higher education by expanding opportunities for academic writing, feedback, tutoring, research support, and personalised learning. This systematic review synthesised empirical evidence on GenAI use in higher education, focusing on the tools employed, reported effects on student learning outcomes, and factors influencing integration. The review followed the PRISMA 2020 framework. Searches of PubMed, Web of Science, and Scopus covered peer-reviewed English-language studies published from January 2021 to January 27, 2026. The search identified 20,671 records; 2,615 remained after database filtering, 188 were sought for retrieval, and 170 full texts were assessed. Joanna Briggs Institute appraisal classified 84 studies as high quality, 74 as medium quality, and 12 as low quality; the low-quality studies were excluded, leaving 158 studies for synthesis. ChatGPT was the most frequently examined tool, appearing in 112 studies (70.9%). Academic writing and essay support was the most common application area (28 studies, 17.7%), followed by personalised tutoring and feedback (22 studies, 13.9%). Seventy-two studies found positive effects on academic performance, while other studies reported non-significant or negative outcomes. Digital and AI literacy findings were predominantly positive, whereas evidence on critical thinking, creativity, and metacognitive skills was mixed. Effective integration was associated with individual readiness, ethical and social considerations, institutional support, pedagogical design, and technical infrastructure. Overall, outcomes depend on purposeful, supported, and critically engaged use. The review highlights the importance of contextualised institutional guidance, professional development, ethical safeguards, and adequate infrastructure for sustainable educational use. These findings suggest that access to GenAI tools alone does not guarantee improved learning; educational value depends on how tools are integrated into pedagogy, assessment, learner support, and critical engagement.