You feel dumber, or less sure of yourself, since you started relying on AI. You are not imagining it.

It is 6pm and you cannot remember the last thing you decided

You close the laptop. It was a busy day. Emails went out, a proposal got sent, a client said thanks.

Then a small question shows up. When did you last really decide something? Not approve. Decide. You sit there and you cannot find it.

You used to look at a messy problem and feel your way to an answer. Now you type it in, read the reply and think "yes, fine, send." It works. That is the strange part.

Nobody else can tell. The work looks good. But you feel like a guest in your own chair.

Why "just trust yourself more" does not help

The usual advice is to have more confidence. Believe in your own thinking. Take breaks from screens.

None of that touches the cause. You are not low on self-belief. You are low on practice. Confidence in your judgment came from using it, over and over, being right or wrong and adjusting. When the AI answers first every time, you skip the step where your own answer forms.

There is a second trap. Offloading feels good. The work goes out faster and the client says thanks. Skip the gym and your body tells you. Skip thinking and the feedback is positive. So nothing tells you to stop.

The book puts it in one line, and I agree with it. You do not disappear into a bad tool, because a bad tool fights you. You disappear into a good one.

What the research does and does not say

A 2025 study in the journal Societies asked 666 people about their AI use and tested their critical thinking. Frequent AI use was linked with lower scores, and the link ran through cognitive offloading, the habit of handing your thinking to the tool (Gerlich, Societies).

Read the word "linked" slowly. This is a link at one point in time, based partly on what people said about themselves. It does not prove AI causes the drop. The book is careful about this, and so should you be.

Here is why it still matters. Judgment is the thing that tells you when AI is wrong, and AI is sometimes wrong in a calm, sure voice. In a study of 758 consultants, people using AI on a task just past its range were about 19% less likely to get it right than people working alone (Dell'Acqua et al., Organization Science). The more you lean on the tool, the duller the one sense you need to catch that.

What about the MIT "Your Brain on ChatGPT" study?

If you searched this feeling, you have probably seen headlines about an MIT study. Here is what it did, in plain terms.

Researchers at the MIT Media Lab had 54 people write essays while their brain activity was measured with EEG (Kosmyna et al., arXiv). One group used ChatGPT, one used a search engine and one used no tools. The ChatGPT group showed the weakest brain connection patterns. They also had trouble quoting from essays they had just written, and felt the least ownership of them. The authors call this "cognitive debt."

Two cautions. It is a preprint, which means it was posted before full peer review, and it is small. Only 18 people did the last session. So treat it as a warning sign, not settled science. The book does not build on it for that reason. It leans on the larger survey of 666 people above.

One detail in it fits this page well. In that last session, people who had first written on their own and then switched to ChatGPT did better on recall than those who had used ChatGPT from the start. Thinking first, then using the tool, is the same order this page suggests.

Do your own thinking first, a few times a week

The shift is small. On calls that matter, take thirty seconds to write your own rough answer before you read the AI's. Where do you think this lands? What would you do?

Then read the AI's answer. Now you are comparing two views instead of taking one. When they agree, you go ahead with more confidence. When they do not, you have found something worth checking.

You cannot do this after you read the AI's reply. Its answer fills the space where yours would have formed. So the order matters.

The book adds one more habit. Once a week, pick one real decision and do it the long way, with the AI off until you have your own answer. It compares this to a pilot who flies by hand now and then, so the autopilot is not the only thing that knows how. Jonah did this and felt his old instinct come back, stiff at first, then strong. He did not use AI less after that. He used it more, and he was in charge of it again.

The book written for this feeling

This is the problem AI Won't Run Your Business. You Will. was written for.

The Jagged Edge: Knowing When the Confident Machine Is Wrong explains why judgment rusts when you stop using it, and gives you three habits to keep it sharp while still using AI for most of your work.

The Person You Become: Don't Disappear Into the Tools is the chapter for the guest-in-your-own-chair feeling. It ends with the Agency Audit, three lines that show where you are still deciding and where you have slipped into approving without thinking.

The free Judgment Workbook has the Hand-on-the-Wheel Check, five questions scored 0 to 3. Your lowest score tells you which chapter to read first. If doubt is the part that feels worst, see using ChatGPT to calm business doubt.