Random Knowledge Verification as a Scalable Assessment Strategy in AI Rich Higher Education

The rapid advancement of generative artificial intelligence (GenAI) is altering student workflows within higher education. Concerns have arisen regarding large language models’ ability to produce articulate essays, reports, and presentation materials. Many institutions have attempted to respond by implementing strategies to identify AI-generated work and increasing the enforcement of academic integrity. In the end, detection methods have not resolved the primary concern of whether students truly comprehend course content. This article posits that the greatest concern with the integration of AI into education is not detecting the use of AI, but rather confirming the presence of learning. The article presents Random Knowledge Verification (RKV) as a practical assessment method that focuses on confirming students’ understanding. The study is a classroom-based analysis conducted over a period of two academic years at a public university in Vietnam with 1,498 undergraduate students across 38 classes in the social sciences, humanities, and communication fields. Students participated in individual and group work, and their performance was evaluated through standard grading and knowledge verification exercises that incorporated technology restrictions, random verbal questioning, and brief written recall activities. The data show a persistent gap between Assignment Quality and Knowledge Mastery, with an average AI-Learning Gap of 2.07 out of 10. Even though students submitted assignments that appeared sophisticated, many struggled to describe or reconstruct the core ideas when other resources were unavailable. Thus, these findings indicate that sophisticated academic outputs, in this case, do not necessarily demonstrate genuine understanding in an AI-augmented learning environment. The study suggests that assessment reform should focus on learning verification rather than AI detection. Additionally, Random Knowledge Verification is a practical, low-cost method for increasing the visibility of learning in AI-rich higher education classrooms.