This is the third post in what has become a series on AI and cognition. In the first, I looked at what happens to neural engagement when students use ChatGPT to write essays (Kosmyna et al., 2025). In the second, I explored whether AI genuinely increases productivity or just makes us work harder (Ranganathan & Ye, 2026). Both posts pointed toward a common thread: the cognitive cost of letting AI handle our thinking.
Gerlich (2025) pulls that thread further. His study examines the relationship between AI tool use and critical thinking across a broad population of users, combining survey data, statistical modeling, and qualitative interviews. The findings are sobering, and they add an important layer to what the neuroscience and workplace research have already been telling us.
AI Cognitive Offloading: The Mechanism Behind the Problem
The central concept in Gerlich’s study is cognitive offloading, the process of delegating memory, evaluation, and decision-making tasks to technology. Every time a user asks AI to summarize a document, evaluate an argument, compare options, or draft a recommendation, they transfer cognitive work that would otherwise require sustained mental effort.
That transfer has consequences. Gerlich found “there is a strong negative correlation, indicating that increased use of AI tools is associated with lower critical thinking skills.” (p. 14).
The relationship holds up even after controlling for education, age, and occupation. This is a significant finding because it means the decline in critical thinking linked to AI use cannot be explained away by demographic differences. The AI reliance itself plays a meaningful role in shaping how people think.
What makes this study particularly valuable is that it identifies AI cognitive offloading as the mechanism driving this decline. Gerlich reports that “The indirect effect through cognitive offloading was also significant (b = −0.25, SE = 0.06, p < 0.001), indicating that cognitive offloading partially mediates this relationship. The direct effect of AI usage remained significant (b = −0.17, SE = 0.05, p < 0.01).” ( p. 14).
In other words, AI doesn’t just correlate with weaker thinking. It actively reshapes how thinking work gets distributed between the human and the machine. Each time the machine takes on more, the human exercises less.
Education Helps, But It’s Not Enough
One finding I found particularly interesting is about education level. Participants with more formal education demonstrated stronger critical thinking overall and showed greater awareness of AI’s limitations. That’s reassuring, but only up to a point.
Even among highly educated users, increased AI use still predicted declines in critical thinking. Education provides some protection, a kind of buffer, but it doesn’t fully counteract the offloading effects. This is worth paying attention to, especially for those of us in higher education who might assume that our students’ academic training will naturally inoculate them against uncritical AI use. The data suggests otherwise.
I covered a related point in a previous post on cognitive surrender, where Shaw and Nave (2026) showed how even well-educated participants defer to AI outputs without critically evaluating them. The pattern across both studies is consistent: knowing that AI can be wrong doesn’t automatically translate into the habit of checking whether it is.
Younger Users Are More Vulnerable
Gerlich’s data also reveals age-related differences that deserve attention. Younger participants showed higher dependence on AI tools and lower engagement in deliberate, reflective approaches to problem-solving. Older participants were more likely to report using careful, independent reasoning processes.
This makes sense when you consider exposure. Younger users have grown up with AI-assisted tools as a normal part of their cognitive environment. The habit of reaching for AI first, before attempting a task independently, becomes ingrained. And if the Kosmyna et al. (2025) brain imaging data is any guide, that habit has measurable consequences for neural engagement. The less you practice independent thinking, the less your brain activates the networks that support it.
The implications for education are hard to ignore. If students develop their cognitive habits during the years when AI cognitive offloading is most available and most tempting, the cumulative effects could be significant. This doesn’t mean keeping AI out of classrooms. It means being deliberate about when and how students engage with it, and making sure independent thinking comes first.
What Users Say About Their Own Thinking
One of the strengths of Gerlich’s study is the qualitative component. Beyond the statistical models, he interviewed AI users about how they experience their own cognitive habits. The responses are candid.
One participant said: “The more I use AI, the less I feel the need to problem-solve on my own. It’s like I’m losing my ability to think critically” (Gerlich, 2025, p. 20).
That’s a remarkable admission. The person recognizes the erosion happening in their own thinking but continues to use AI because the convenience is hard to give up. Several participants described a similar tension: they appreciate the speed and ease AI provides, but they sense that something is being lost in the process. Analytical depth, problem-solving confidence, and the willingness to sit with a difficult question are all diminishing.
As Gerlich summarizes: “These findings collectively demonstrate that high AI tool usage negatively impacts critical thinking, primarily mediated through cognitive offloading” (2025, p. 19).
AI Cognitive Offloading and What Comes Next
This study, combined with the Kosmyna brain imaging research and the HBR work intensification findings, paints a picture that the education and professional worlds need to take seriously. AI is a powerful tool. It can handle tasks that are genuinely better suited to automation. But when it takes over the cognitive work that builds and maintains our thinking capacity, the long-term costs outweigh the short-term gains.
The answer, as I’ve argued across this series, comes down to intention and structure. Let AI handle the mechanical, the repetitive, the logistical. Protect the thinking, the reasoning, the evaluation. Build habits that put independent effort first and AI support second. And keep questioning whether the convenience AI provides is worth the cognitive capacity it quietly takes away.
Reference
- Gerlich, M. (2025). AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), Article 6. https://doi.org/10.3390/soc15010006
- Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing tasks. MIT Media Lab. https://www.media.mit.edu/publications/your-brain-on-chatgpt/
- Ranganathan, A., & Ye, X. M. (2026, February 9). AI doesn’t reduce work—it intensifies it. Harvard Business Review. https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies-it
- Shaw, S. D., & Nave, G. (2026). Thinking fast, slow, and artificial: How AI is reshaping human reasoning and the rise of cognitive surrender. Working paper, The Wharton School, University of Pennsylvania. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6097646