AI in education: boosting productivity or challenging cognitive effort?

Purpose Artificial intelligence (AI) is reshaping higher education by changing how students learn and perform academic tasks. Although AI is associated with greater time efficiency and improved academic performance, its broader effects on student productivity and cognitive effort remain insufficiently understood. This study aims to examine how university students associate AI-led productivity gains with less study time, better grades, better understanding of study material, changes in cognitive effort they put into AI-supported task completion and dependency on AI tools. Design/methodology/approach The authors conducted an exploratory survey among university students in Hanoi in late 2024 and proposed an empirical model that pools self-reported time efficiency, academic performance, academic self-efficacy, change in cognitive effort and AI dependency as dimensions of student productivity. The authors control for AI use intensity and AI use in collaboration, as well as for some other socio-demographic characteristics. Given the exploratory nature of the study and the relatively small sample, the empirical analysis follows a staged specification strategy when the authors first estimate a parsimonious baseline model and subsequently introduce two sets of background controls sequentially. The authors further estimate an interaction between cognitive effort change and AI dependency, first in a parsimonious specification and subsequently with additional controls. Finally, the authors produce a sensitivity analysis of the model to alternative operationalisation of self-efficacy variable. Findings Students associate their AI-led productivity gains primarily with less study time, a bit less so with better grades and better understanding of the study material. They also associate the productivity growth with greater changes of cognitive effort they put into problem-solving while using AI and rising dependency on AI tools. Self-confidence appears to be associated with likelihood to reap significant rather than some productivity gains particularly once the interaction of AI dependency and AI-led cognitive shift is accounted for. The survey data also support the existence of potentially uneven academic benefits from AI use when higher-achieving students are more likely to report further improvements in academic performance, whereas students with weaker prior performance report rather limited gains. Research limitations/implications The findings highlight the need for research design that captures the dynamic evolution of AI-supported learning. The observed differences between daily and weekly users, particularly in critical evaluation and perceived accuracy, suggest that future studies should examine how evaluative skills and cognitive effort develop over time. Longitudinal and experimental designs, combined with objective performance and AI-usage data, would allow researchers to test whether the productivity dimensions proposed here operate causally or how much the fast-paced evolution of AI tools changes the cognitive effort students put into problem-solving and critical thinking when completing academic tasks, and their dependency on AI tools for academic productivity. Practical implications The results provide actionable guidance for universities seeking to integrate AI into teaching and assessment practices. The strong role of time efficiency and self-efficacy, alongside uneven benefits across student groups, indicates the need for structured AI literacy modules that teach source verification and evaluative strategies and effective prompt formulation. Given that daily users show higher critical evaluation but lower perceived accuracy, teaching practices should incorporate reflective assignments that require students to justify, critique and revise AI-generated content. Assessment formats may need redesign to ensure transparent AI use while still requiring independent reasoning, analytical justification and domain-specific understanding. Social implications Policies should balance the productivity benefits of AI-supported learning with safeguards that promote independent thinking, academic integrity and equitable access. Addressing disparities in AI literacy and access is essential to prevent the widening of the existing academic inequalities. Originality/value This study contributes to the emerging literature on AI-assisted learning by examining students’ perceived productivity gains alongside several concurrent dimensions of their AI-supported academic experience, including time efficiency, perceived academic performance, academic self-efficacy, changes in cognitive effort and AI dependency. Rather than proposing these dimensions as new constructs or causal mechanisms, this study brings them together within a single analytical framework. Particular attention is given to whether perceived productivity gains and growing reliance on AI coexist with self-reported changes in cognitive effort, an interplay that remains insufficiently explored in the existing research. Empirically, this study provides exploratory evidence from Vietnam, an underrepresented higher education context in Southeast Asia.