As AI tools spread through classrooms and offices, new research raises concerns that heavy reliance on AI prompts may weaken critical thinking and learning skills.
 |
| Research suggests generative AI improves results but may reduce mental engagement, forcing educators and employers to rethink how AI should be used. Image: CH |
London, United Kingdom — December 21, 2025:
The explosive growth of generative AI has turned simple prompts into a daily cognitive shortcut for millions. From students drafting essays to professionals analysing data, tools like ChatGPT and Copilot are increasingly embedded in how people think, work, and learn. But as AI-driven productivity rises, a more unsettling question is gaining traction: are these tools reshaping human thinking in ways that may ultimately weaken it?
Recent research suggests the concern is not unfounded. A study by the Massachusetts Institute of Technology found that students who used ChatGPT to write essays showed reduced activity in brain networks associated with cognitive processing. More strikingly, they struggled to recall or explain their own arguments afterward. The implication is not just about academic integrity, but about diminished mental ownership of ideas when AI performs core thinking tasks.
Similar patterns appear beyond the classroom. Research conducted by Carnegie Mellon University in collaboration with Microsoft found that workers who expressed higher confidence in AI tools applied less critical thinking to tasks. When AI was trusted to “get it right,” human oversight declined. The risk, researchers warned, is long-term overreliance that erodes independent problem-solving skills.
Yet the picture is far from one-dimensional. Surveys of UK schoolchildren by Oxford University Press reveal a paradox: while six in ten students felt AI had harmed aspects of their schoolwork skills, nine in ten also reported that AI helped them develop at least one ability, such as revision techniques, creativity, or problem-solving. The data points to a crucial variable—not whether AI is used, but how.
Experts argue that the distinction lies between AI as a cognitive partner and AI as a cognitive substitute. Professor Wayne Holmes of University College London warns that students may achieve higher grades with AI-assisted work while learning less in the process. The outcome, he says, is “better outputs but worse learning,” a trade-off that undermines the fundamental purpose of education.
This phenomenon echoes concerns in other professions. In medicine, AI-assisted radiology has improved diagnostic accuracy for some clinicians while degrading skills for others, a form of cognitive atrophy that occurs when humans defer too readily to automated judgment. The lesson, researchers argue, is that AI must be designed and used to keep humans actively engaged, not mentally sidelined.
Technology companies acknowledge the tension. OpenAI, which operates ChatGPT, has emphasized that its tools are not meant to replace thinking but to guide it, promoting features that encourage step-by-step reasoning and interactive learning. Still, critics argue that without clear standards, training, and independent evidence, responsibility is being pushed onto users who may not fully understand how these systems work.
The debate ultimately exposes a deeper challenge: generative AI is not just another efficiency tool, like spellcheck or calculators, but a system that can mimic reasoning itself. That makes it uniquely powerful—and uniquely risky. Whether AI prompts weaken thinking skills or enhance them may depend less on the technology and more on whether societies put guardrails around its use.
As AI adoption accelerates, the question is no longer if people will rely on it, but whether they will remain cognitively present while doing so. Without deliberate effort, the convenience of AI may come at a cost that only becomes visible when critical thinking is most needed.