AI doesn't make you dumber. Using it to skip your thinking does. The difference comes down to one question: are you asking AI to think for you, or to think with you?

The worry is real. A 2025 study of 666 people by Michael Gerlich found that heavier AI users scored lower on critical thinking, and that the link ran through "cognitive offloading": handing mental work to a tool instead of doing it yourself (PsyPost). The study shows correlation, not cause, but the pattern is hard to ignore.

A Microsoft Research survey of 319 knowledge workers adds a twist. The more people trusted the AI, the less critical thinking they applied. The more they trusted themselves, the more they questioned what it produced (The Decoder).

Researchers now draw a useful line between two habits. Habitual offloading uses AI as a replacement for thinking. Strategic offloading uses it as a thinking partner while you keep the wheel (Journal of Intelligence, 2026).

This article is about the second habit. Here are three ways to use AI that make your mind sharper, not softer.

1. Use AI to challenge your assumptions

Every plan, opinion and decision rests on assumptions you've stopped noticing. AI is a tireless sparring partner for dragging them into the light, as long as you ask it to push back instead of agree.

By default, most AI assistants lean agreeable. Ask "Is my idea good?" and you'll often get polite encouragement. Ask "What would have to be true for my idea to fail?" and you get something you can actually use.

How to do it:

  • Name your position first. Write down what you believe and why, in your own words, before you open the chat. This keeps the thinking yours.
  • Ask for the strongest counterargument. Not a strawman. The case a smart, informed opponent would make.
  • Ask what you're taking for granted. List the hidden premises your argument depends on, then test each one.
  • Argue back. Don't accept the critique wholesale either. Defend, concede or refine. The back-and-forth is where the thinking happens.

Prompts to try:

  • "Here's my view and my reasoning. Steelman the opposing position as persuasively as you can."
  • "List the assumptions my argument depends on. Rank them by how shaky they are."
  • "Play a skeptical expert in this field. What would you challenge first?"
  • "Don't reassure me. Tell me where my logic is weakest."

The goal isn't to change your mind every time. It's to hold your views because they survived scrutiny, not because they never met any.

2. Use AI to uncover your blind spots

Assumptions are things you believe without checking. Blind spots are worse: things you never thought to consider at all. You can't question what you can't see, but AI can help you widen the frame.

We all view problems through the lens of our role, our industry, our experience and our mood that day. AI has absorbed an enormous range of perspectives, which makes it good at asking "Have you considered...?" about angles outside your usual field of view.

How to do it:

  • Swap the lens. Ask how the situation looks to someone with different stakes: a customer, a competitor, a new hire, a regulator, your future self.
  • Run a pre-mortem. Imagine it's a year from now and your plan failed. Ask AI to write the most likely story of why.
  • Look for what's missing. Ask which questions, data or stakeholders your thinking doesn't account for.
  • Check your patterns. Share how you made a past decision and ask which known biases might have shaped it: confirmation bias, sunk cost, overconfidence.

Prompts to try:

  • "Here's my plan. Who would be affected by it that I haven't mentioned, and how might they react?"
  • "It's 12 months from now and this project failed. Write the three most likely explanations."
  • "What important questions am I not asking about this?"
  • "Describe this situation from the point of view of someone who strongly disagrees with my priorities."

Treat the answers as leads, not verdicts. Some of what AI raises will be irrelevant. Deciding which blind spots matter is the part of the work that stays with you, and it's the part that builds judgment.

3. Use AI to deepen your thinking process

The first two uses improve what you think. This one improves how you think. Instead of asking AI for answers, ask it for better questions, and do the answering yourself.

This is the Socratic method on demand. A good teacher rarely hands you the solution. They ask the question that makes you see the next step on your own. The struggle is the point: it's how understanding sticks.

How to do it:

  • Think first, then ask. Spend 10 minutes with a blank page before you open the chat. Your rough draft gives the AI something to sharpen instead of replace.
  • Ask it to question you, not answer you. Tell it explicitly to hold back conclusions and ask one probing question at a time.
  • Explain it back. After learning something, explain it in your own words and ask AI to find the gaps in your explanation.
  • Make your reasoning visible. Ask it to help you map your argument step by step, so you can see where a leap in logic hides.

Prompts to try:

  • "I'm working through this problem. Don't give me the answer. Ask me questions that help me get there."
  • "Here's my explanation of this concept. What did I get wrong or leave out?"
  • "Walk me through how an expert would break this problem down, then let me try the next one myself."
  • "What's a question about this topic that would push my understanding one level deeper?"

You'll get to the answer more slowly this way. That's the trade. You'll also remember it, understand it and be able to do it again without the tool.

Bonus: Assume AI can't read your mind, and get specific

AI is artificial. It doesn't know your goals, your audience, your constraints or what "good" looks like to you. Assume it's bad at inferring what you mean, and you'll be forced to spell it out. That act of spelling it out is a thinking exercise in itself.

Vague requests get generic answers. "Help me with my marketing plan" gets boilerplate. "Help me find the weakest assumption in a plan to reach 200 small-bakery owners in Quebec with a $2,000 budget" gets something useful, and you had to get clear on your own plan just to write it.

How to do it:

  • State the context. Who it's for, what it's for, and what you've already tried.
  • Name the constraints. Budget, time, length, tone, what's off the table.
  • Define success. Describe what a great answer looks like before you ask for one.
  • Notice where you struggle. If you can't fill in a detail, that's not an AI problem. It's a gap in your own thinking worth closing first.

Prompts to try:

  • "Before you answer, ask me the questions you'd need to give a truly specific response."
  • "What details am I leaving out that would change your answer?"

Over time, this habit carries beyond the chat window. You'll start writing clearer briefs, asking sharper questions in meetings and catching fuzzy thinking in yourself before anyone else does.

The rule of thumb: think first, ask second

All three uses share one principle. You bring the thinking; AI pressure-tests it. When the order flips, when AI produces the thinking and you just approve it, that's when the skills start to fade.

Before your next prompt, run a quick check:

  1. Have I formed my own view first?
  2. Am I asking AI to challenge, widen or question my thinking, or to replace it?
  3. Will I understand this better afterward, or just have an answer?

Used this way, AI isn't a crutch. It's a gym. And like any gym, it only works if you're the one lifting.

Sources

Ideas mine, English polished with AI. How I use AI →