The Side Effects of Over Reliance on AI

Posted Mar 6, 2026

By David Ayrton 6 min read

Making a Case

Artificial intelligence tools have become deeply embedded in how we work, study, and think. From ChatGPT writing our emails to GitHub Copilot completing our code, the convenience is undeniable. But convenience has a cost — and a growing body of research suggests that leaning too heavily on AI may be quietly eroding the very cognitive skills that make us effective in the first place.

What the Research Says

Critical Thinking is Taking a Hit

A 2025 study by Michael Gerlich at SBS Swiss Business School surveyed 666 participants across diverse age groups and found a significant negative correlation between frequent AI tool usage and critical thinking scores. The mechanism? Cognitive offloading — the tendency to outsource mental effort to an external system. Cognitive offloading correlated strongly with AI tool usage (r = +0.72) and inversely with critical thinking ability (r = −0.75). Younger users aged 17–25 were the most affected.

A separate study from Microsoft and Carnegie Mellon University reinforced this, finding that the more knowledge workers trusted AI-generated outputs across 936 tasks, the less cognitive effort they applied overall.

Source: Increased AI use linked to eroding critical thinking skills — Phys.org

Test Scores and Learning Retention

Research conducted by the University of Pennsylvania with Turkish high school students found a troubling split: students using ChatGPT answered 48% more problems correctly during practice — but scored 17% lower on a concept understanding test afterward. AI boosted procedural performance while undermining deeper comprehension.

An MIT Media Lab study tracked students writing essays with ChatGPT-4 over four months. LLM users consistently underperformed at neural, linguistic, and behavioral levels compared to those who didn’t use AI. 83% of students in the AI group couldn’t recall key points from their own essays. The group relying exclusively on ChatGPT also showed the least brainwave activity during writing tasks.

Sources:

Cognitive Atrophy: A Long-Term Risk

Neurologists and researchers have raised concerns about cognitive atrophy — the gradual weakening of mental faculties from underuse. Just as muscles weaken without exercise, cognitive pathways for reasoning, problem-solving, and memory consolidation can deteriorate when AI handles those tasks consistently. A review published in Smart Learning Environments (Springer) identified this pattern across multiple studies: over-reliance on AI dialogue systems measurably reduced students’ cognitive abilities over time.

Polytechnique Insights frames it clearly: generative AI carries a real risk of cognitive atrophy, particularly for those who have not yet achieved mastery in their field — which includes most students and early-career professionals.

Sources:

AI in Professional Work: Pros and Cons

Pros

Speed on routine tasks. AI excels at generating boilerplate code, summarizing documents, drafting templates, and handling repetitive work. Teams can redirect that time toward higher-order thinking.
Broad adoption signals real value. The 2025 Stack Overflow Developer Survey found that 84% of developers are using or planning to use AI tools, with 51% using them daily. Around 81% of GitHub Copilot users say it helps them complete tasks faster.
Accessibility. AI lowers the barrier to entry for complex tasks — junior developers can produce working code, non-writers can draft polished copy, and non-designers can prototype UI concepts.

Cons

It can actually make you slower. A rigorous randomized controlled trial (METR, 2025) found that experienced developers using AI tools took 19% longer to complete tasks than those working without them — the opposite of the expected outcome.
Code quality and tech debt. AI code assistants can rapidly generate “almost right” code that accumulates silently into technical debt. Debugging AI-generated code was cited as the top frustration among developers in the Stack Overflow survey.
Accuracy and trust issues. 87% of respondents in the same survey expressed concern about AI accuracy, and 81% flagged security and privacy risks.
Declining enthusiasm. Positive sentiment toward AI tools dropped from over 70% in 2023–2024 to 60% in 2025 — a sign that real-world friction is tempering early optimism.

Sources:

Finding the Balance: Learning With AI vs. Letting AI Do It All

The goal isn’t to avoid AI — it’s to use it in a way that builds your capabilities rather than replacing them. Here are some practical techniques:

1. Try it yourself first

Before prompting an AI for a solution, spend 10–20 minutes working through the problem independently. This activates recall and strengthens the neural pathways that AI would otherwise bypass. Research on pretesting shows that students who attempted problems before using AI retained significantly more than those who went straight to the tool.

2. Use AI to check, not to generate

Write the first draft — of code, of an essay, of a plan — yourself. Then use AI to review, critique, or identify gaps. This keeps you in the driver’s seat cognitively while still benefiting from AI’s breadth.

3. Explain AI outputs back to yourself

When AI gives you an answer, close the window and explain it in your own words. If you can’t, you don’t understand it yet. This technique, rooted in the Feynman method, forces genuine comprehension rather than passive acceptance.

4. Set AI-free zones

Designate certain tasks or time blocks as AI-free. For developers, this might mean solving one algorithm problem a day without assistance. For writers, it might mean drafting one piece per week from scratch. Deliberate practice without assistance keeps skills sharp.

5. Use AI for exploration, not execution on unfamiliar topics

When learning something new, AI is excellent for generating questions, providing overviews, and suggesting directions — but let yourself do the actual work of understanding. Use it as a map, not a car.

6. Reflect on what you actually learned

After completing an AI-assisted task, ask yourself: what did I learn from this? If the answer is “nothing,” that’s a signal to pull back and re-engage more directly next time.

The Bottom Line

AI is a powerful tool, and like any powerful tool, it rewards skilled, intentional use and punishes passive dependency. The research is consistent: moderate, deliberate AI use can enhance performance, but excessive reliance measurably degrades critical thinking, memory, and learning outcomes — especially for those still building expertise.

The professionals who thrive long-term won’t be the ones who use AI the most. They’ll be the ones who use it wisely — augmenting their judgment rather than replacing it. Be kind and do good.

This post is licensed under CC BY 4.0 by the author.