Abstract
The rapid expansion of artificial intelligence (AI) tools in higher education has reopened fundamental questions about the trajectory of students' critical thinking (CT) development. This article reports the results of a longitudinal cohort diagnostic study conducted at three higher education institutions in Astana, Republic of Kazakhstan L.N. Gumilyov Eurasian National University, Kazakh National University of Sports, and Saken Seifullin Kazakh Agro Technical Research University across four measurement points: 2020, 2022, 2024, and 2026, with a total sample of 334 students. CT was assessed using the Starkey Critical Thinking Test, adapted for Russian-language audiences, which measures analytical reasoning, argument evaluation, fallacy identification, evidence assessment, and problem-solving capacity. Inferential analyses employed the Mann–Whitney U test, Cohen's d effect sizes, and the Kruskal–Wallis omnibus test. Three research hypotheses were formulated and tested: that CT levels would demonstrate a statistically significant positive trajectory over the study period; that the period of active AI adoption (2023–2024) would coincide with a peak in CT performance; and that subsequent normalisation of AI use would be associated with a partial decline. The results confirm statistically significant improvement from 2020 (M = 14.07, SD = 3.71) to 2024 (M = 18.58, SD = 4.44), with a large effect size (d = 1.10, p < 0.001), and a statistically non-significant partial decline in 2026 (M = 17.03, p = 0.059). The overall Kruskal–Wallis test confirms significant between-cohort differences (H = 43.13, p < 0.001). The findings are discussed in relation to active pedagogical strategies, media literacy, and the dual role of AI as both a stimulus and a potential inhibitor of critical reasoning. Implications for competency-based curriculum reform within the Bologna Process framework are considered.