Generative AI in Southeast Asian TVET: A Systematic Review of Learning Outcomes, Ethics and Research Directions

Background: Generative artificial intelligence tools such as ChatGPT entered technical and vocational education and training faster than evidence on their effects. Objective: This review synthesized empirical evidence on the applications, learning outcomes and ethical challenges of generative artificial intelligence in Southeast Asian vocational education. Method: Following PRISMA 2020, ten search strings were run in OpenAlex and Semantic Scholar for 2022–2026, which yielded 254 unique records, of which 25 primary studies met the eligibility criteria. Findings: The studies came from Indonesia (17), Malaysia (5) and Thailand (3), and 18 were published in 2026. Fourteen concerned English language learning. Four quasi-experimental studies reported gains in writing, speaking, writing self-efficacy and AI literacy, while most other evidence was self-reported, and no study measured hands-on technical competence. Academic integrity (9 studies), overreliance (7) and inaccurate output (5) were the most frequent risks, and privacy and unequal access were also reported. Benefits depended on teacher mediation, task design, AI literacy and institutional guidance. Implications: Institutions should favors guided over unrestricted use, and research should prioritize controlled, longitudinal studies with practical assessment. Contribution: The review offers the first region-specific evidence map and a framework for responsible integration.