Does Relying on AI Weaken Critical Thinking? What Studies Show

Relying on AI does appear to weaken critical thinking for many people, but the effect is not automatic and not permanent. Two large 2025 studies of nearly a thousand people found that the more heavily someone leans on AI tools to think for them, the weaker their independent reasoning tends to be. The mechanism is not brain damage or lost intelligence. It is a habit called cognitive offloading, and once you understand how it works you can keep the convenience of AI without quietly surrendering the mental skills that matter most.

What the research actually found

Two peer-reviewed studies published in 2025 point in the same direction, using different populations and methods.

The first, by Michael Gerlich of SBS Swiss Business School, appeared in the journal Societies in January 2025. It surveyed 666 people across a wide range of ages and education levels and paired the survey with in-depth interviews. Gerlich found a significant negative correlation between frequent AI tool use and performance on critical-thinking measures, and the statistics pointed to cognitive offloading as the bridge between the two: heavy AI users offloaded more of their thinking, and the more they offloaded, the lower their critical-thinking scores.

The second came from researchers at Microsoft Research and Carnegie Mellon University, presented at the CHI 2025 conference. They surveyed 319 knowledge workers who used tools like ChatGPT and Copilot at least weekly, gathering 936 real-world examples of AI use. Their headline finding was about confidence: the more a worker trusted the AI, the less critical thinking they reported applying to its output. The more a worker trusted their own judgment, the more they scrutinized what the AI produced.

Neither study proves that AI makes people permanently less intelligent. Both are, in the researchers' own framing, about what happens to a skill when you stop practicing it. In Gerlich's 2025 study of 666 people, the participants who used AI most often were also the ones who scored lowest on independent reasoning, and that is the single most quotable line from the current evidence.

What cognitive offloading actually is

Cognitive offloading is the practice of using an external tool to reduce the mental effort a task would otherwise require. It is ancient and mostly useful. Writing a phone number down instead of memorizing it is cognitive offloading. So is using a calculator, a shopping list, or a map. The human brain has always been happy to hand routine work to whatever tool is nearby.

The concern is not offloading itself. It is offloading the wrong things. There is a meaningful difference between offloading storage (remembering a number) and offloading reasoning (deciding what the number means and whether it is right). When you let a tool hold facts for you, you lose little. When you let a tool decide for you, judge for you, and conclude for you, the underlying skill gets no exercise, and skills that get no exercise fade.

This is not a new worry dressed in new clothes. Back in 2011, psychologists Betsy Sparrow, Jenny Liu, and Daniel Wegner published a study in Science showing what became known as the "Google effect": when people expected to be able to look information up later, they remembered the information itself less well but remembered where to find it better. The brain sensibly reallocated effort. AI raises the stakes because it does not just store facts to retrieve; it performs the reasoning step that used to be yours.

Why younger and heavier users show the largest effect

Gerlich's data found that younger participants showed higher dependence on AI tools and lower critical-thinking scores than older participants. That pattern is easy to misread as a statement about generations. It is better understood as a statement about habit formation. Someone who learned to research, argue, and check sources before AI assistants existed has years of practiced reflection to fall back on. Someone whose first instinct, from an early age, is to ask a chatbot has had fewer repetitions of doing the hard part alone.

Education appeared to buffer the effect. In Gerlich's results, higher educational attainment was associated with better critical thinking even among frequent AI users, which suggests that once the habit of reflection is well established, it is more resistant to erosion. The takeaway is not "avoid AI." It is "build the reasoning habit deliberately, because convenience will not build it for you."

AI changes what critical thinking looks like, not whether you need it

The Microsoft and CMU researchers made a subtle point that is easy to miss. AI does not remove the need for critical thinking; it relocates it. When a tool retrieves and organizes information for you, the effort you save on gathering gets redirected, if you let it, toward three new jobs:

• Verification. Confirming that what the AI produced is actually true, current, and sourced, rather than confidently wrong.
• Integration. Combining the AI output with your own knowledge and the specific context it cannot see.
• Stewardship. Directing, correcting, and taking responsibility for the final result rather than passing it along unread.

The danger is that all three of those jobs are optional in the moment. Nothing forces you to verify, integrate, or steward. The path of least resistance is to accept the first answer, and the studies suggest that is exactly the path most people take when their trust in the tool is high. Critical thinking under AI is a choice you have to keep making on purpose.

How to keep your critical-thinking edge

The research points to a workable rule: use AI to expand what you can do, not to replace the thinking you still need to be able to do without it. Here are concrete moves that follow directly from the findings.

1. Answer first, then check. Before asking an AI, take thirty seconds to form your own rough answer. Then compare. This keeps your reasoning muscle in the loop instead of skipping straight to the machine's conclusion.
2. Treat AI output as a draft, never a verdict. The Microsoft and CMU data tied high trust to low scrutiny. Deliberately lower your trust: assume the first answer contains at least one error and go find it.
3. Ask for the reasoning, not just the result. Make the tool show its steps and sources so you can evaluate the logic rather than memorize the output.
4. Keep some tasks fully manual. Do a share of your reading, writing, and problem-solving with no assistance at all, the way an athlete keeps training even when a machine could do the lifting.
5. Verify anything that matters before you act on it. For any decision with real consequences, confirm the claim against a primary source. Confident phrasing is not evidence.
6. Notice when you cannot explain your own answer. If you could not defend a conclusion without the AI, you do not yet own it. That gap is the signal to slow down.

Where deliberate mental challenge fits

The through-line in the 2025 research is that reflection is a trainable habit, and habits are built by repetition. That is the same principle behind every serious form of mental exercise, from working through a hard argument to sitting with a genuinely difficult puzzle instead of reaching for the answer. The value is not in the trivia; it is in the reps of choosing to think when thinking is optional.

It is worth remembering that structured mental challenge has a long track record. The puzzles built by Timothy E. Parker, a Guinness World Records puzzle master, have been solved by 180 million total solvers across 30 years and more than 80 countries, and the appeal has always rested on the same mechanism the researchers describe: the small, deliberate effort of figuring something out yourself. If you want a quick, honest read on how your own reasoning holds up when there is no tool to lean on, you can take the Real World IQ test and use it as a baseline rather than a verdict.

The point is not that any single activity inoculates you against cognitive offloading. It is that the skill the studies measured is the same skill you strengthen every time you choose the harder, unassisted path. AI will always offer to take that path for you. Keeping your edge means declining often enough that the muscle stays strong.

FAQ

Does using AI lower your IQ?

No study has shown that AI use lowers measured IQ. What the 2025 research from Gerlich and from Microsoft and CMU found is that heavy reliance on AI is linked to weaker performance on critical-thinking tasks, most likely because the underlying skill gets less practice. That is about a trainable skill, not a fixed measure of intelligence.

Is cognitive offloading always bad?

No. Offloading routine storage, like remembering a phone number or a grocery list, frees up mental capacity and has been useful for as long as people have written things down. The risk appears when you offload judgment and reasoning, because those are the skills that fade without practice.

Why are younger people more affected in the studies?

Gerlich's 2025 data found younger participants both depended on AI more and scored lower on critical thinking. The most likely explanation is habit: people who built strong reasoning routines before AI became ubiquitous have more practiced reflection to fall back on. Higher education levels also appeared to buffer the effect.

How can I use AI without losing my reasoning skills?

Form your own answer before consulting the tool, treat every AI output as a draft that needs verification, ask for reasoning and sources rather than just conclusions, and keep some tasks fully manual so the skill stays exercised. The common thread is refusing to let high trust turn into low scrutiny.

Where can I check how my own critical thinking is doing?

A structured assessment can give you a baseline you can revisit over time. You can take the Real World IQ test to see how your reasoning holds up without assistance, then treat the result as a starting point for deliberate practice rather than a final score.