Topic and Problem Statement

Artificial intelligence (AI) has garnered significant attention in vocational education and training (VET), resulting in a rapidly expanding body of review literature synthesising empirical findings on AI applications, opportunities, and challenges. However, this landscape is characterised by fragmentation. Existing reviews vary markedly in scope - ranging from technical analyses of adaptive learning systems to broad surveys of emerging technologies - and often lack theoretical integration (Fatokun & Gumbo, 2025; Ghosh & Ravichandran, 2024). Consequently, although the volume of literature continues to increase, the knowledge base remains disjointed, offering limited cumulative understanding of how AI interacts with the institutional logics that shape vocational systems.

Objective and Research Questions

The objective of this study is to conduct a qualitative meta-review of review literature on AI in VET. By synthesising review-level evidence, the study moves beyond descriptive summaries to develop an integrative understanding of how AI is conceptualised and evaluated within vocational contexts. The inquiry is guided by three research questions:

1. What themes and focal areas dominate the review literature on AI in VET?
2. Where do existing reviews converge and where do they diverge regarding the impact and implementation of AI?
3. What conceptual or theoretical gaps emerge from a cross-review synthesis, particularly in relation to system-level integration?

Theoretical Framework

To address the observed fragmentation, the study adopts a Three-Level Framework derived from the Multi-Level Perspective on socio-technical transitions developed by Geels (2002). This framework organises the evidence across three analytical dimensions:

1. Technological Functionality (Niche Level): AI is conceptualised as a set of tools and innovations, such as generative AI for instructional content creation and adaptive assessment systems (Ranuharja et al., 2025).
2. Pedagogical and Organisational Embedding (Regime Level): AI adoption is viewed as embedded within institutional cultures, professional practices, and organisational routines. This level emphasises organisational readiness, leadership, and teacher adaptation as mediating factors (Rahayu, 2025).
3. System-Level Governance (Landscape Level): AI use is shaped by regulatory frameworks, qualification standards, and broader policy environments that define legitimate practices and accountability structures (Geels, 2002; Rabiu et al., 2025).

The framework structures the identification of dominant themes (RQ1) and patterns of convergence and divergence (RQ2) across levels. It also supports the development of explanatory mechanisms linking technological potential to educational outcomes: didactic mediation, organisational translation, and governance framing.

International Dimension

The study places strong emphasis on the international dimension of VET. Comparative evidence highlights that AI effectiveness is highly context-dependent. Vocational institutions in resource-constrained settings are particularly vulnerable to widening inequalities due to infrastructure deficits and policy vacuums. This perspective is essential for understanding divergent AI adoption pathways between the Global North and Global South (Gao & Tan, 2025; Rabiu et al., 2025).

This study employs a qualitative meta-review design (review of reviews). Unlike a quantitative meta-analysis, it does not aggregate primary data statistically. Instead, it synthesises review-level findings to identify cross-cutting patterns, conceptual tensions, and theoretical gaps across a heterogeneous body of literature. Corpus Selection The corpus consists of review publications explicitly addressing AI in VET, TVET, or vocational higher education. A broad search strategy was used to ensure comprehensive coverage. Searches were conducted in curated databases (Scopus and Web of Science) as well as crawler-based academic search engines (Google Scholar and Semantic Scholar). The final corpus includes 26 review studies, encompassing systematic reviews (with and without PRISMA) and also narrative reviews. Analytic Approach Data analysis followed a second-order thematic synthesis based on the method developed by Thomas and Harden (2008). Instead of primary studies, the analytic units were the findings, results, and discussion sections of the included review articles. The process comprised three stages: 1. Line-by-line coding of review findings to capture key meanings and concepts. 2. Development of descriptive themes through clustering related codes across studies. 3. Generation of analytical themes that extend beyond the original reviews to develop higher-order interpretations aligned with the three-level framework.

The meta-review identifies strong convergence in framing AI as a pedagogical support technology, particularly in relation to adaptive learning and automated feedback. However, effectiveness is consistently described as conditional rather than inherent. Educational outcomes are mediated by: Organisational Readiness
: Successful implementation depends on leadership, infrastructure, and a supportive digital workplace culture. Teacher adaptation is shaped by technology acceptance and change readiness (Rahayu, 2025). Contextual Vulnerabilities: 
VET systems are especially susceptible to equity risks. Without targeted investment, AI adoption may exacerbate the digital divide in resource-constrained contexts (Gao & Tan, 2025). The analysis also reveals persistent structural tensions, including trade-offs between efficiency and pedagogical judgement, and between personalisation and the standardisation embedded in vocational qualification frameworks. Additionally, tensions arise between rapid technological innovation and slower-moving governance systems, limiting scalability when regulatory frameworks lag behind institutional experimentation (Rabiu et al., 2025). Implications The primary challenge in this field is not the absence of empirical research, but the lack of theoretical integration. Future research should adopt cross-level designs that connect classroom applications with organisational and governance structures. For policymakers, the findings underscore the importance of robust ethical and regulatory frameworks to prevent policy vacuums and ensure equitable access to AI-enhanced vocational education and training (Rabiu et al., 2025; Solak Berigel et al., 2025).

Fatokun, J. O., & Gumbo, M. T. (2025). A systematic review of the impact of emerging technologies in shaping technical and vocational education and training students’ learning styles. Educational Challenges, 30(2). https://doi.org/10.34142/2709-7986.2025.30.2.17 Gao, H., & Tan, Y. (2025). AI differences in vocational and undergraduate education: Differential applications of artificial intelligence in undergraduate and vocational higher education: A systematic review. Higher Education Studies, 15(4), 398. https://doi.org/10.5539/hes.v15n4p398 Geels, F. W. (2002). Technological transitions as evolutionary reconfiguration processes: A multi-level perspective and a case-study. Research Policy, 31(8-9), 1257–1274. https://doi.org/10.1016/S0048-7333(02)00062-8 Ghosh, L., & Ravichandran, R. (2024). Emerging technologies in vocational education and training. Journal of Digital Learning and Education, 4(1), 41-49. https://doi.org/10.52562/jdle.v4i1.975 Hamdan, A., Elmunsyah, H., Maula, P. I., Sari, H. V., Abdullah, M. F., & Septianingsih, D. (2025). Systematic evaluation of adaptive assessment in e-learning: Contribution to vocational education. Proceedings of the 9th International Conference on Electrical, Electronics and Information Engineering (ICEEIE), 1-6. https://doi.org/10.1109/ICEEIE66203.2025.11254863 Mustakim, W., Giatman, Maksum, H., Abdullah, R., & Refdinal. (2024). Integration of policy and technology in vocational education leadership: A systematic literature review. Jurnal Pendidikan Teknologi Informasi dan Vokasional. https://doi.org/10.23960/jptiv.v6i2.33016 Rabiu, A., Bawa, K., & Saminu, S. (2025). Artificial intelligence adoption for skills development in Nigeria: A systematic review and roadmap for TVET transformation. International Journal of Research and Innovation in Social Science, IX(VIII). https://doi.org/10.47772/IJRISS.2025.908000051 Rahayu, S. (2025). AI readiness, virtual leadership, and digital workplace culture in vocational education: A systematic review on how technology acceptance and change readiness shape future teaching skills. Proceedings of the International Conference on Economic Business Management, and Accounting (ICOEMA) 2025. https://e-conf.usd.ac.id/index.php/icre/ICRE2025/paper/download/5569/924 Ranuharja, F., Ganefri, Rizal, F., Langeveldt, D., Ejjami, R., Torres-Toukoumidis, A., & Jalinus, N. (2025). Relevance and impact of generative AI in vocational instructional material design: A systematic literature review. Salud, Ciencia y Tecnología, 5, 1336. https://doi.org/10.56294/saludcyt20251336 Solak Berigel, D., Şilbir, L., & Şilbir, G. M. (2025). Integrating artificial intelligence (AI) into technical and vocational education and training (TVET): A PRISMA-based systematic review. Calitatea Vieții, 36(1), 83-108. https://doi.org/10.46841/RCV.2025.01.05 Thomas, J., & Harden, A. (2008). Methods for the thematic synthesis of qualitative research in systematic reviews. BMC Medical Research Methodology, 8(45), 1–10. https://doi.org/10.1186/1471-2288-8-45