Does Generative AI Improve Academic Performance? Evidence from Higher Education in Kazakhstan

The public release of generative artificial intelligence (AI) chatbots in late 2022 has been widely recognized as reshaping how university students study. However, rigorous evidence on whether AI use translates into academic gains remains limited, particularly outside North America and Western Europe. To our knowledge, this is among the first empirical studies examining the relationship between generative AI use and academic performance in higher education in Central Asia. This study examines the relationship between AI tool availability, AI usage intensity, measured by the share of assignments completed with AI, and grade point average (GPA) among 110 undergraduates at Nazarbayev University in Kazakhstan. The study combines a cohort comparison based on AI availability with a cross-sectional analysis of AI usage frequency, assignment coverage, and self-reported dependence. To maximize statistical power, two university entrance examinations (SAT and NUET) were merged into a standardized measure of prior academic ability. After controlling for standardized prior academic ability, gender, study hours, and major, none of the three AI usage measures, nor cohort-level AI availability, is significantly associated with GPA. Prior academic ability remains by far the strongest predictor of current GPA. The findings remain consistent across multiple robustness checks. Exploratory analysis suggests two patterns of further investigation: a possible concave (diminishing-returns) relationship between AI-assisted assignment coverage and GPA, and an association that is significant only among students who took the NUET exam. These findings contribute to the evidence that access to generative AI alone does not necessarily lead to improved academic performance and highlight the need for larger, multi - institutional studies to better understand when, how, and for whom AI supports learning.