AI-driven assessment in higher education: a systematic review

The rapid advancement of artificial intelligence (AI) is reshaping the concepts and practices of assessment in higher education. Despite the growing body of empirical research, there remains a lack of systematic understanding of the developmental trends, core functional applications, and broader impacts of AI-driven assessment in this field. To address this gap, this study employed the PRISMA framework to systematically review 172 empirical studies published between 2010 and 2025 across five major international databases. The findings reveal that: (1) research on AI-driven assessment is characterized by a strong predominance of quantitative approaches and a marked concentration in STEM disciplines; (2) AI technologies have been applied primarily to assessment design and delivery, automated grading and feedback, and learning analytics and performance monitoring; and (3) AI-driven assessment shows considerable potential for improving assessment efficiency, supporting personalized learning, informing instructional decision-making, and optimizing assessment systems, while also facing substantial challenges related to technical performance, psychometric quality, algorithmic transparency, ethics, and governance. These findings support a conceptual understanding of AI-driven assessment as an augmentative rather than substitutive tool, and yield practical guidance for aligning technology, pedagogy, and governance to realize its educational value in higher education.