Key Takeaways
- AI can harm critical thinking by replacing the mental struggles necessary for learning.
- Using AI can lead to cognitive offloading, where people remember less because they expect the information to be available elsewhere.
- AI can create a false sense of confidence, leading users to accept incorrect information as true because of its authoritative tone.
- While AI can diminish ownership in creative work, it can also enhance learning when used as a guiding tutor rather than an answer machine.
- To maintain critical thinking skills, use AI after you’ve tried to solve a problem on your own and done deep cognitive work.
Artificial intelligence (AI) is making millions of people faster. Debate is growing over whether AI weakens thinking by making tasks too easy.
It can summarize a report, draft an email, explain a difficult topic, analyze data, and generate 50 ideas before your coffee gets cold.
That speed feels like a superpower.
If AI always thinks first, you will eventually struggle to think without it.
The threat is not that ChatGPT will suddenly erase your intelligence. The threat is that you will slowly stop practicing the skills AI performs for you.
You may write less, remember less, question less, and accept more.
At the same time, telling people to avoid AI is unrealistic and shortsighted.
For example, calculators did not destroy mathematics, and search engines did not end human knowledge.
The better question is not:
“Does using AI hurt our ability to think?”
Instead, it is:
“Which ways of using AI weakens my thinking, and which ways does it strengthen it?”
The early evidence offers a complicated but useful answer.
How AI quietly weakens your thinking
AI does not automatically make you less intelligent, but using it as an answer machine can slowly replace the mental effort that helps you learn. Let’s dive in.
1. AI can remove the mental struggle that produces learning
Learning is not supposed to feel effortless.
You learn when you retrieve information, connect ideas, test possible answers, discover gaps, and correct mistakes.
That struggle is not a flaw in the process.
That struggle is the process.
AI can remove that struggle in seconds.
When you ask an AI tool to explain a concept before attempting to understand it yourself, you receive a polished answer without building the mental path that leads to it.
The answer may enter your document without entering your memory.
Long before generative AI, learning researchers studied desirable difficulties or challenges that can slow performance during practice while improving long-term learning.
Research from UCLA psychologists Robert and Elizabeth Bjork shows why effortful practices such as retrieval, spacing, and varied practice can strengthen retention.
Easy performance is not the same as durable learning.
You may feel productive because AI finished the task. However, the quality of the output doesn’t prove you developed the underlying ability.
A finished answer can hide an unfinished mind.
Before asking AI, spend 5-10 minutes thinking alone.
- Write what you know.
- Identify what confuses you.
- Attempt an answer.
Then bring in the tool.
“The important thing is not to stop questioning.”— Albert Einstein, the Nobel Prize-winning physicist who transformed our understanding of space and time
2. AI can make you remember less
Your brain is efficient.
When it expects information to remain available elsewhere, it may remember where to find it rather than the information itself.
Psychologists observed this years before ChatGPT arrived.
A landmark study led by Columbia University psychologist Betsy Sparrow found that people were less likely to remember information when they believed it would be saved. The research suggested that internet access was changing what people remembered.
AI takes cognitive offloading one step further.
Search engines help you find information. Generative AI can find, organize, explain, interpret, and rewrite it for you.
That convenience helps when the information is unimportant. Nobody needs to memorize every airport code, spreadsheet formula, or software command.
The risk appears when you outsource the knowledge you need to think.
You cannot evaluate an AI-generated argument about economics, history, health, or business strategy if you have no knowledge of your own. Critical thinking does not happen in an empty head.
You need facts in your mind to recognize when AI is inventing them.
After using AI to learn something, close the tool and explain the idea from memory. If you cannot explain it simply, you have not learned it yet.
You have only borrowed an answer.
3. AI can reduce your sense of ownership
Think about the last time you wrote something entirely on your own.
You probably remember wrestling with the opening, changing your position, removing weak points, and discovering what you believed as you wrote.
That friction creates ownership.
A 2025 MIT Media Lab experiment compared people writing essays with ChatGPT, traditional search, or no digital tools. The research reported lower neural engagement, weaker recall, and less ownership among the ChatGPT group.
This study deserves caution.
It involved a small sample, focused on one type of essay task, and was released as a preprint rather than a final peer-reviewed paper. It does not demonstrate that normal AI use permanently damages the brain.
Still, its central concern is worth taking seriously.
When a tool creates the structure, language, examples, and conclusion, it becomes difficult to tell where its thinking ends and yours begins.
You may submit excellent work and feel strangely disconnected from it.
If you cannot defend an idea without reopening the AI conversation, the idea was never fully yours.
Write your first position before asking AI for its position. After using the tool, remove anything you cannot explain or defend independently.
“I think, therefore I am.”— René Descartes, the French philosopher and mathematician whose work shaped modern Western philosophy
AI can produce text. It cannot take responsibility for what you believe.
4. AI can make confident nonsense feel true
AI-generated answers often sound calm, organized, and certain.
That tone is dangerous because humans tend to confuse confidence with competence.
A detailed explanation feels more trustworthy than a hesitant one, even when the confident answer is wrong.
Researchers call our tendency to accept computer guidance “automation bias.”
Research has shown that automated systems can create new biases rather than simply removing human ones.
Generative AI raises the stakes because it does not merely recommend an answer.
It can create the evidence, logic, quotation, and professional language needed to make the answer appear credible. A hallucination can arrive wearing a very nice suit.
The answer also becomes easier to accept when it agrees with you.
You prompt the tool with your assumptions, receive a response shaped by those assumptions, and then treat the response as independent support. That is confirmation bias with better formatting.
“The first principle is that you must not fool yourself — and you are the easiest person to fool.”— Richard Feynman, the Nobel Prize-winning physicist known for his curiosity and intellectual honesty
Harvard researchers tested interventions that forced people to think before seeing or accepting AI advice.
Their research found that these cognitive forcing techniques reduced overreliance on incorrect AI recommendations.
The catch was revealing: people liked the safer systems less.
We enjoy tools that remove friction, even when the friction protects us.
Convenience feels good because it asks less of you. Good judgment sometimes requires doing the inconvenient thing.
Before accepting an important AI answer, ask:
- What evidence supports this?
- Which assumptions shaped the response?
- What could make this conclusion wrong?
- Can I verify the claim with a primary source?
- What would a smart critic say?
Never outsource the final judgment when the consequences belong to you.
5. AI can improve thinking when it acts like a tutor
So far, this may sound like an argument against AI. It isn’t.
AI can improve learning when it guides your thinking instead of replacing it.
In a randomized controlled trial, Harvard students using a carefully designed AI tutor learned significantly more in less time than students in an active-learning classroom.
Research also found that the AI group reported greater engagement and motivation.
The important detail is easy to miss.
The tutor didn’t just hand students finished answers. It used structured prompts, scaffolding, feedback, and questions to help them work through the material.
That is a different relationship with AI.
An answer machine ends the thinking process. A tutor keeps it moving.
Try prompting AI this way:
“Help me understand this topic, but don’t give me the answer immediately. Ask one question at a time, challenge weak reasoning, and make me explain each step.”
You can also ask it to quiz you, provide a counterargument, identify missing evidence, or adjust the difficulty based on your answers.
The best AI doesn’t think for you. It creates conditions that make you think better.
Bringing it all together
Does using AI hurt our ability to think? AI weakens our thinking.
If you use AI to avoid effort, skip uncertainty, generate your first thought, remember everything for you, and deliver conclusions you never examine, your thinking skills may weaken.
That outcome is not mysterious. Any ability declines when you stop practicing it.
But AI can also help you learn faster, encounter new perspectives, test assumptions, receive personalized feedback, and improve weak reasoning. The same tool that allows intellectual laziness can support intellectual growth.
The difference is how you use it.
AI is not making everyone less intelligent. AI is making passive thinkers more dependent and active thinkers more capable.
AI will widen the gap between people who want answers and people who want understanding.
Use it after you think, not before. Ask it for questions, not only answers.
Make your first attempt without help. Verify important claims with original sources, and never publish an idea you cannot explain in your own words.
Most importantly, keep doing difficult cognitive work.
- Read long articles.
- Write from a blank page.
- Solve problems without assistance.
- Sit with confusion long enough to learn what it is trying to teach you.
Your brain stays sharp not because you own intelligent tools. It stays sharp because you keep using it. AI can quietly weaken your thinking if you don’t know how to stop it.
Join my email newsletter for a free eBook and more helpful insights.