Introduction

The rapid adoption of AI tools such as ChatGPT, Claude, and Gemini has prompted a significant debate among researchers, educators, and professionals: are these tools making us cognitively weaker? Recent studies have attracted considerable attention, and some have used their findings to make the claim that AI is making people less intelligent. This essay argues that while the underlying research raises legitimate and important concerns, the blanket conclusion misrepresents a more nuanced and conditional reality. The evidence suggests that the cognitive impact of AI depends critically on how it is used, who is using it, and the degree to which users remain actively engaged in their own thinking.

What the Research Actually Shows

The most widely cited study comes from MIT, which examined three groups of students completing essay-writing tasks using, respectively, no technology, Google (specifically its traditional search engine, not its AI capabilities), and ChatGPT-4. The results were striking: students who exclusively used ChatGPT demonstrated the least brainwave activity of any group, showed measurable decreases in cognitive function in key areas over time, and exhibited the weakest connection to their own work. A notable 83% could not recall key points from essays they had just written (MIT, 2024).

A 2025 study by Swiss researcher Michael Gerlich similarly found that frequent reliance on AI tools can cause critical thinking abilities to atrophy, a phenomenon he termed "cognitive offloading." When we consistently delegate intellectual tasks to an external system, the neural pathways associated with those tasks receive less activation and, over time, may weaken, much as a muscle weakens through disuse.

These findings are serious and deserve attention. They are not, however, a basis for concluding that AI inherently diminishes human intelligence.

The Critical Distinction: Replacement vs. Augmentation

The MIT study measured the effects of exclusive AI use, specifically students who surrendered the entire cognitive task to the tool. This is a fundamentally different behaviour from using AI as a thinking partner, a research accelerator, or a critical sounding board. The research does not show that AI assistance in general causes cognitive decline; it shows that outsourcing thinking entirely does.

This distinction mirrors well-established patterns with other cognitive tools. GPS navigation, when used as a complete substitute for spatial reasoning, is associated with weaker navigation skills. Yet no serious researcher would conclude that maps make people less capable. The concern is always about the manner and degree of use, not the tool itself.

Who Is Using AI Matters Enormously

A March 2026 analysis in Psychology Today makes a compelling and underappreciated point: the effect of AI use is likely categorically different depending on the user's existing skill level. An experienced professional using AI to summarise a report they already know how to evaluate is offloading a task they have already mastered, and the benefit is efficiency. A young student using AI to write essays before they have developed the underlying skills risks never building those capacities at all.

This is a meaningful distinction that the blunter framing of AI as a straightforward cognitive threat entirely obscures. The concern is most acute in educational contexts involving younger users, and least applicable to experienced adults using AI as a professional tool. It is precisely for this reason that AI use in education must be carefully considered and implemented with a clear understanding of how and when it is appropriate, ensuring that it supports the development of skills rather than bypassing it altogether.

The Case for Thoughtful Optimism

It is worth noting that using AI well is itself a cognitively demanding skill. Formulating precise prompts, critically evaluating AI output, identifying errors and biases, and knowing when not to trust the tool all require active, sophisticated thinking. Users who engage with AI in this way are not cognitively passive; they are operating at a higher level of abstraction. Research from Microsoft supports this: a study comparing different modes of AI assistance found that a system which encouraged users to articulate their own reasoning before receiving AI feedback helped them examine their thought processes and consider alternative perspectives more deeply than one which simply provided ready-made answers (Microsoft Research, 2025). In other words, the design of the interaction, and the intentionality of the user, shapes the cognitive outcome entirely.

Furthermore, access to a knowledgeable, patient, and responsive thinking partner was historically limited to those with significant social or financial capital. AI has the potential to democratise that access, enabling more people to explore ideas, receive feedback, and develop their thinking than ever before. A 2026 longitudinal study by Gonsalves found that students who engaged iteratively and reflectively with AI tools showed evidence of strategic thinking, adaptive learning, and metacognitive growth, moving fluidly between cognitive and reflective domains in ways that deepened rather than diminished their understanding (Gonsalves, 2026). Whether this potential is more broadly realised depends entirely on how the tool is used.

When AI Actively Enhances Cognitive Performance

The positive evidence is substantial. A review of over 5,000 studies found that human-AI groups outperform humans working alone in 85% of cases, with professional writing quality rising by 18% and time taken falling by 40% (Vaccaro et al., 2024). A survey of 5,000 knowledge workers found that strategic AI collaborators were 1.8 times more likely to be seen as innovative, with gains in both work quality and motivation (Atlassian, 2025). A systematic review of critical thinking literature found that active AI engagement supports broader perspective-taking, stronger argumentation, and deeper understanding of complex material (Melisa et al., 2025). Across all three bodies of research, the consistent finding is the same: the cognitive benefit of AI scales with the quality of human engagement.

Conclusion

The claim that AI is diminishing human cognition is not wholly without foundation, but it is stated far too broadly to be treated as a general truth. The research identifies real risks associated with a specific pattern of behaviour: passive, uncritical, wholesale delegation of thinking to AI systems. Those risks are most serious for users who have not yet developed the underlying cognitive skills being offloaded.

For users who engage with AI actively, questioning its outputs, using it to challenge and extend their own thinking, and retaining ownership of their reasoning, the evidence does not support a conclusion of cognitive decline. The determining variable is not the tool, but the intentionality of the person using it.

What the current body of research ultimately demonstrates is that the relationship between AI and cognition is neither straightforwardly harmful nor straightforwardly beneficial. It is conditional, contextual, and shaped by human behaviour. That conclusion places the emphasis squarely where it belongs: not on the technology, but on the choices we make in using it.

The more useful question, then, is not whether AI poses a risk, but how individuals, institutions, and educators can establish the conditions under which it enhances rather than undermines human thinking. The research points not to a reason to avoid AI, but to a responsibility to use it wisely.

References

Gerlich, M. (2025). AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), 6. https://doi.org/10.3390/soc15010006

MIT Media Lab. (2024). Study on cognitive engagement and LLM use in essay writing tasks. Massachusetts Institute of Technology. https://www.nextgov.com/artificial-intelligence/2025/07/new-mit-study-suggests-too-much-ai-use-could-increase-cognitive-decline/406521/

Psychology Today. (2026, March 22). Adults lose skills to AI. Children never build them. The Algorithmic Mind. https://www.psychologytoday.com/us/blog/the-algorithmic-mind/202603/adults-lose-skills-to-ai-children-never-build-them

Loeb, A. (2026). Harvard professor on cognitive decline from AI use. Futurism. https://futurism.com/artificial-intelligence/harvard-avi-loeb-ai

The Conversation. (2026, January 22). Is AI hurting your ability to think? How to reclaim your brain. https://theconversation.com/is-ai-hurting-your-ability-to-think-how-to-reclaim-your-brain-272834

Atlassian. (2025). AI collaboration report: Using AI is not enough. https://www.atlassian.com/blog/productivity/ai-collaboration-report

Gonsalves, C. (2026). Generative AI Impact on Critical Thinking: Revisiting Blooms Taxonomy. Journal of Marketing Education. https://journals.sagepub.com/doi/10.1177/02734753241305980

Melisa et al. (2025). Critical thinking in the age of AI: A systematic review. EDUPIJ. https://files.eric.ed.gov/fulltext/EJ1459623.pdf

Microsoft Research. (2025). The future of AI in knowledge work: Tools for thought at CHI 2025. https://www.microsoft.com/en-us/research/blog/the-future-of-ai-in-knowledge-work-tools-for-thought-at-chi-2025/

Vaccaro et al. (2024), cited in: Zhang, Z. T. et al. (2025). Augmenting human cognition with generative AI: Lessons from AI-assisted decision-making. arXiv. https://arxiv.org/abs/2504.03207