What If AI Could Train Your Brain Instead of Thinking for You?

You’re asking AI to reassure you — not to challenge you

Here’s what I notice, in myself as much as in most people who use AI every day.

You ask it a question. You present it with an idea. And, without realizing it, you’re waiting for one thing only: for it to confirm what you already thought.

“Does this approach seem good to you?”
“Do you agree with me?”
“Does that sound right?”

These are closed questions. They only call for one answer: yes.

When your question is already framed to get a validation, what you mostly get back is an answer that stays within the frame you set yourself.

Result: you walk away from the exchange reassured. But not any further ahead.

Substitution or stimulation: the real question AI raises about your reasoning

There’s a real difference, from the standpoint of your reasoning, between two ways of using AI.

Substitution. You hand over the thinking. It thinks, you approve. You become a spectator of your own reasoning.

Stimulation. You ask it to question, to push back, to offer a counter-argument you wouldn’t have seen coming. You stay the one doing the thinking.

This isn’t just intuition. It’s also what researchers are starting to document.

A 2025 study by researcher Michael Gerlich surveyed 666 people on their daily AI habits and tested their critical thinking scores. According to the study, the more often people use AI, the lower their scores on tasks that require reasoning and reflection.

The explanation he offers is easy to follow: the more you hand a piece of your mental effort over to an outside tool, the less you train yourself to do it. Researchers call this cognitive offloading.

One important nuance: this study shows a link between the two, not proof that AI alone causes a drop in critical thinking. It’s a snapshot at a given moment, not a demonstration of cause and effect.

A more recent study, led in 2026 by researcher Qiuhan Zhu and her team (Frontiers in Psychology), takes this further on one essential point: it suggests that not all cognitive offloading is created equal.

The researchers distinguish two ways of relying on AI. In the first, the person hands over a significant share of their own involvement in the reasoning — AI takes the lead. In the second, AI stays a support: the person keeps confronting what it produces with their own thinking. These are, roughly, what I call here substitution and stimulation — worth noting these are the words I use to make the distinction easy to grasp, not the researchers’ own terminology.

Across 589 students and early-career professionals followed over several weeks, the “substitution” mode was associated with lower intrinsic motivation and less favorable perceived outcomes on creativity, independent judgment, and deep reflection. The “stimulation” mode, on the other hand, preserved motivation and came with more favorable perceived outcomes.

One point stands out to me in particular: in the short term, both modes deliver a comparable benefit. That’s exactly where the trap lies. Delegating your thinking to AI can be very efficient in the moment — a fast answer, less effort, a feeling of productivity.

The real question isn’t what AI let you produce today. It’s what you, personally, keep exercising or not.

These findings rely largely on what participants perceive and self-report, not on an objective measure of brain activity. They raise an important question more than they settle a debate.

So the difference isn’t in the tool. It’s in how you use it.

Why it’s so tempting to be told you’re right

Several distinct mechanisms explain why we usually prefer confirmation over contradiction.

The first is cognitive economy. The brain relies on automatic patterns, habits, and shortcuts so it doesn’t have to consciously re-evaluate every piece of information that comes its way. Like every other organ, it tries to limit its energy spend — and it already uses a lot: barely 2% of your body weight, but close to 20% of your energy.

The second is confirmation bias. We naturally tend to seek out, interpret, and remember information that confirms what we already believe — and to downplay what contradicts it.

The third is what a real contradiction actually costs: it forces you to reopen a line of reasoning you thought was settled, and to tolerate the discomfort of being wrong or revising your position.

These three mechanisms work together. And when you stay locked inside your own thinking, never confronting it with an outside perspective, you eventually mistake your own stories for reality.

This is exactly where AI can play a different role than the one we usually give it.

Not a tool that thinks for you.

A tool that helps you stress-test your reasoning, spot your blind spots, and build that precious capacity: metacognition. It’s what lets you avoid automatically accepting AI’s first answer — and instead ask where it comes from, and whether it really holds up. This capacity to observe your own thoughts is exactly what I share about on this topic.

Turning AI into an intellectual sparring partner

A sparring partner, originally, is a training partner. Someone who trades blows with an athlete or a professional to help them improve, to prepare for competition.

They don’t fight to win. They fight to make you better.

The term has widely made its way into the business world. It’s now used for an advisor, a mentor, an outside consultant.

Their role isn’t to do the work for you. They challenge your ideas, push you out of your comfort zone, and help you step back from your strategic decisions. It’s a role I take on often with the executives and managers I work with, at key decision-making moments.

That’s exactly the role AI can play with your thinking — as long as you ask it to, clearly.

5 ways to use AI without outsourcing your brain

This won’t be a list of magic prompts.

These are five cognitive reflexes. Each one paired with a concrete question to ask AI.

1. Write your own answer before asking for its answer

Principle: produce before comparing.

Before you ask AI anything, write your own answer first. Even imperfect. Even incomplete.

This one simple habit changes everything. You do the work first. You’re not just receiving a ready-made answer — you’ve already built one of your own.

Example: “Here’s my answer: […]. Now tell me what you’d change.”

2. Ask it to find the flaws in your reasoning

Principle: confrontation.

Instead of asking whether your idea holds up, ask what could make it fall apart.

Example: “What are the three biggest weaknesses in this reasoning?”

3. Ask for a position opposite to yours

Principle: cognitive flexibility.

Reasoning that only ever explores one point of view eventually gets mistaken for the truth. Ask AI to seriously defend the opposite of what you think.

Example: “Argue seriously for the opposite of my position.”

4. Question the answer instead of just accepting it

Principle: metacognition, critical thinking.

An AI’s answer isn’t a truth. It’s a proposal. Ask it what it’s basing that on, and where the limits of what it just said actually are.

Example: “What are you basing that on? What are the limits of this answer?”

5. Keep the final decision

Principle: cognitive autonomy.

AI can inform, argue, challenge. It should never decide for you. You’re the one who makes the call — with more to go on, but with your own judgment.

Example: “List the arguments for and against. The final decision will be mine.”

Taking back the power over your thinking

I don’t see AI as a threat to your intelligence.

The real question isn’t how many tasks you can hand over to it. It’s which abilities you want to keep exercising yourself.

Analyze. Doubt. Challenge. Verify. Choose.

AI can generate a huge number of answers, very fast. That doesn’t mean it should exempt you from building your own judgment.

A crutch does part of the effort for you. A training partner, on the other hand, pushes you to produce your own.

That’s the posture I invite you to adopt with AI — rather than enduring it or avoiding it.

Next time you open a conversation with it, try something different. Don’t ask if you’re right.

Ask it to prove you wrong.

And if you’d like to go further in understanding these mechanisms — how your brain works by default, and how to consciously retrain it — you’ll find practical tools in my book NeuroExcellence.

Sources

Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.

Zhu, Q., Li, X., Dong, Y., Chang, P., & Fan, M. (2026). Not all cognitive offloading is equal: distinguishing dependent and autonomous offloading to generative AI. Frontiers in Psychology, 17, 1878629.