DOI: https://doi.org/10.36347/sjet.2026.v14i10.005

Pages: 644-653

Purpose: Generative artificial intelligence (AI) has changed the economics of social engineering. Large language models write fluent, personalised phishing at near-zero cost, voice-cloning tools imitate a lecturer or parent from seconds of audio, and real-time deepfake video can impersonate an official on a call. Universities combine an open culture, high staff and student turnover, large financial flows and young, heavily connected users, which makes them attractive targets. Yet the literature on AI-enabled deception and the literature on cybersecurity in higher education have developed largely in parallel. This review brings them together. Method: An integrative review following Whittemore and Knafl's (2005) approach. Structured searches of Scopus, Web of Science, ScienceDirect, SpringerLink, IEEE Xplore and the ACM Digital Library (2017–2026), supplemented by citation chasing, identified empirical studies, systematic reviews and authoritative reports on AI-enabled social engineering, human deepfake detection, phishing susceptibility and literacy-based interventions. The evidence was synthesised thematically and mapped onto a conceptual model. Findings: Four findings stand out. First, generative AI amplifies six threat vectors relevant to universities, from large language model (LLM) spear-phishing to synthetic-persona scholarship and romance fraud. Second, pooled human deepfake detection accuracy is close to chance (about 55% across 56 papers), and people consistently overestimate their own ability. Third, awareness training and media-literacy interventions produce real but modest and decaying gains, and they work better against misinformation than against targeted fraud. Fourth, research on students is concentrated in high-income countries, with very little from sub-Saharan Africa despite the region's specific exposure to financial fraud. Contribution: We propose a Layered Human–Technical Defence Model that links threat vectors, human detection limits, individua