Does Outsourcing Thinking to AI Make Me Dumber?

I use AI for most of my work, I teach it to other people, and I have nine kids watching me do both. Here is what I actually think it is doing to us.

A question landed in my inbox last week, and before I had thought about it for even three seconds, my hand was already opening the AI tab.

I caught myself mid-reach. Not with dread. With curiosity. When did thinking stop being the first thing I do?

I am not a neuroscientist. I run a marketing agency, a wedding venue, and a small AI consulting practice, and somewhere in there I have nine kids spread across elementary school, junior high, high school, and college. I use AI for the majority of my work now. I also teach it to other people, which is the part that forces me to actually understand the machine instead of just poking at it. And the more I teach it, the more I keep circling back to the same uncomfortable question that everybody is quietly asking and almost nobody is answering honestly.

Is this making us dumber?

I want to tell you where I land before I show you the work, because I think the honest answer matters more than the dramatic one. I am in the glass-half-full camp. I think AI is going to be good for us. But I did not get there by ignoring the scary part. I got there by looking straight at it.

The Question Worth Asking Is Bigger Than Me

The small version of this question is about me. Am I getting lazy? Am I losing an edge I was proud of?

The bigger version, the one I think actually connects with people, is about all of us at once. What does handing our thinking to a machine do to how a whole society remembers, learns, and reasons? Not to one knowledge worker having a slow Tuesday. To the population.

I have an unusual seat for this. Four of them, actually. I watch my kids use AI on their homework, and they are at wildly different stages of learning to think. I watch my teams and my clients absorb it into their work. I watch my own habits shift, day by day. And I watch rooms full of professionals when I teach them, which is where the tells are the loudest.

So I did what I do now with any real question. I pointed AI at it and told it to go find the truth. Then I did the harder part. I checked its work.

The Honest Science, Both Hands On the Table

Here is the first thing I found, and it is the thing that keeps this from being a panic piece. This is not new. We have been offloading our thinking to machines for a long time.

Back in 2011, researchers at Columbia published a study in Science on what they called the Google effect. When people believed a fact was saved somewhere they could look it up, they remembered the fact worse, but they remembered where to find it better. Our brains quietly reassigned the job. We stopped storing the information and started storing the address. That was fifteen years ago, with a search bar. AI is the same move, one level up. We are not just offloading memory anymore. We are starting to offload the reasoning.

And there is real evidence that the reasoning part matters more. A 2025 preprint out of MIT put people in three groups to write essays, one using ChatGPT, one using a search engine, one using nothing but their own brain, and measured their brain activity with EEG. The pattern was clean. The brain-only group showed the strongest, most connected mental networks. The search group was in the middle. The AI group showed the weakest connectivity, and afterward they struggled to accurately quote the essays they had just written. They produced the work without quite owning it.

I want to be careful here, because being careful is the whole point. That MIT study is a preprint. It has not been peer reviewed, the sample was small, and other researchers have already raised questions about it. You may have seen headlines claiming "83% of ChatGPT users couldn't quote their own essay." That number is not in the actual paper. It came from commentary and traveled faster than the truth did. This is exactly the trap the article is about, so I am not going to fall into it while writing it.

But it does not stand alone. Microsoft and Carnegie Mellon surveyed 319 knowledge workers in a peer-reviewed study and found something quieter and more unsettling. The more people trusted the AI, the less critical thinking they did. The more they trusted themselves, the more they did. And a study of 666 people published in the journal Societies found a strong negative link between heavy AI use and critical thinking scores, with the strongest effect in 17 to 25 year-olds.

The more you trust the machine, the less you check it. The more you trust yourself, the more you do. That single sentence is the whole ballgame.

If I stopped there, you would close this tab feeling terrible. So I am not going to stop there, because the other hand is just as real.

Offloading is a trade, not a theft. Writing something down is offloading. A calculator is offloading. The research on this is decades old, and it is clear that handing a task to an external tool frees up mental room to do bigger things. The cost is that you remember the offloaded thing less well later. That is a trade. The only question that has ever mattered is whether you chose the trade or sleepwalked into it.

And when AI is aimed well, it does not shrink people. It lifts them. In a study of more than 5,000 customer support agents, access to an AI assistant raised performance by 14% on average. But the average hides the good part. The least experienced agents got 34% better. The most experienced ones barely moved. The AI worked by taking the hard-won instincts of the best people and quietly handing them to everyone else.

AI barely helped the top performers. It made the strugglers 34% better. We keep assuming AI makes the smart smarter. Mostly, it lifts the floor.

Even in learning, the story flips when the design is right. Harvard ran an experiment with 194 physics students. The ones tutored by a well-built AI learned roughly twice as much as the ones in a normal active-learning class, in less time, and they enjoyed it more. The catch, and it is the entire catch, is that the AI was built to ask them questions, not just hand them answers. Same technology. Opposite outcome. The design decided everything.

What COVID Taught Me About How Fast We Change

Here is the mental model that pulled all of this together for me.

Think about what COVID did to how we connect. Not the disease. The habits it left behind. We text instead of call. We take the Zoom even when the drive is easy. We are, on the whole, a little more tucked in than we used to be.

The numbers are stark. The US Surgeon General reported that the time an average American spends in person with friends fell from about 60 minutes a day in 2003 to about 20 minutes a day by 2020. For young people it dropped nearly 70%. And here is the part that matters most. Nobody felt it happen. There was no Tuesday where you woke up and decided to see your friends two-thirds less. The default just drifted, quietly, until one day you looked up and it had calcified.

That is my fear about AI and thinking. Not a dramatic collapse. A quiet drift you do not feel in real time, until you look up in a few years and notice the muscle is not what it was.

But COVID taught me the hopeful half too, and it is the half I actually believe. That drift was not destiny. Remote work settled at about a quarter of paid workdays and stopped falling, and the people who stayed intentional about their relationships came out of it just fine. The default drifted. The deliberate chose. Same fork.

COVID happened to us. This, we get to choose. That difference is everything.

So Here Is What I Actually Predict

I want to label this clearly. This is my projection, not a fact. But I have four front-row seats and I am willing to say what I see coming.

I do not think AI is going to make us dumber on average. I think it is going to widen the gap. The people who use it passively, who let it think so they do not have to, will drift. The people who use it intentionally, who argue with it and check it and use it to sharpen their own reasoning, will pull further ahead than they ever could have before. The thinking does not disappear. It moves. It moves from having the answer to interrogating the answer, and that turns out to be a higher-order skill, not a lower one.

The reason I am hopeful is that the protective factor is learnable. In that study of 666 people, the ones who stayed sharp were not the ones who avoided AI. They were the ones who engaged with it critically. That is a teachable habit. It is the most teachable thing there is.

I will be fair about the other side, because you should not trust me if I am not. This could break the wrong way. If we build tools that just hand out answers, and we raise a generation to accept them without a fight, the drift wins. That is a real risk and I am not going to pretend it away. But if I have to bet, and I do this for a living, I bet on us. I think more people rise than fall. I think the floor comes up faster than the ceiling comes down.

What I See From My Four Seats

This is the part only I can tell you, because it is not in any study.

A few months ago one of my high-schoolers handed me a paragraph for a class assignment that was, honestly, better than anything I wrote at that age. Clean. Confident. Airtight. So I asked one simple question. Walk me through why you argued it that way. And there was nothing there. Not a wrong answer. No answer. The words were on the page, but the thinking had never happened in their head.

Here is what I did not do. I did not ban the tool. We sat down and I had them use the same AI to argue the opposite side, hard, and then defend their original point against it. Twenty minutes later they understood the material better than that polished paragraph ever showed. Same tool. The first way made them look smart. The second way made them think.

I see the same split at Rocket Media almost every week. Two people, same tools, opposite trajectories. One started running everything through AI and the output doubled overnight, but it all began to sound the same. Competent. Generic. Forgettable. The work stopped having a person in it. Someone else picked up the exact same tools and got noticeably sharper, and the only difference was what they asked the machine to do. The first used AI to produce. The second wrote their own take first, then told the AI to tear it apart. One outsourced the thinking. The other hired a sparring partner. Six months in, you can tell those two apart with your eyes closed.

And I watch it happen in almost every room I teach. A hand goes up, and the question is some version of, just give me the prompt that does my job for me. They want the machine to take the thinking off their plate. Then, in the same room, someone asks a completely different question. How do I get it to challenge me? To tell me where I am wrong? Those two people are going to end up in very different places, and it has almost nothing to do with how smart they were when they walked in. It comes down to whether they wanted a crutch or a coach.

The pattern across all three seats is the same. It is never the tool that decides. It is whether the person stays in the chair.

The Part Where I Have to Be Honest About Myself

Which brings me back to that flinch, the one where my hand was on the AI tab before my brain was in gear.

It was a proposal for a client I actually care about. My hand was moving before my brain was, already typing "write me an opening for a proposal about," before I had spent a single second deciding what I wanted to say. So I stopped. I closed the tab. I sat with the blank page, which felt genuinely uncomfortable, longer than I wanted to, and I made myself answer one question. What does this client actually need to hear from me? Then I wrote three rough sentences of my own. Clumsy ones. But mine.

Then I brought the AI back in, not to think for me, but to sharpen what I already had. The version we sent started from my three ugly sentences, cleaned up. It was better than anything the machine would have handed me cold, because it started from something true.

I am the guy who teaches this. If I can drift, anyone can. But that is not the discouraging part. The encouraging part is that I noticed, and noticing was enough to change it. The muscle was still there. It just needed me to choose to use it.

What To Do Tonight

I am not going to tell you to use AI less. I use it more than ever, and I am not sorry about it. I am going to tell you to keep your hand on the wheel, and there is one small habit that does most of the work.

Before you reach for AI, give yourself ten seconds to form your own answer first. Just ten seconds. Then let the machine challenge it, extend it, or prove you wrong. And when you can, make the AI ask you questions instead of only answering yours. That one shift, from a thing that hands you conclusions to a thing that sharpens your own, is the entire difference between the version of this that makes us dumber and the version that makes us sharper.

That is it. Ten seconds of your own thinking before the machine gets a vote. It is small enough to actually do, and it is the whole game.

Where This Really Lands

I keep coming back to my kids at the dinner table, all nine of them, at every stage of learning to think at once.

My oldest learned to reason before any of this existed. My youngest never will. He is going to grow up in a house where the answer to any question is always one sentence away, and the most important thing I can teach him is not how to get the answer. It is when to distrust it. When to sit with a hard question a little longer than is comfortable. When to think first and reach second.

The kids who learn to interrogate the machine are going to run circles around the ones who just obey it. That is not a threat. That is the most hopeful thing I know about this whole moment. The advantage is still ours to take. It was never going to be handed to us, and it was never going to be taken from us. We get to choose.

I am choosing to think first. I hope you will too.

AI is not making us dumber. Unexamined AI use is. And the fix was never fear. It is intention, which happens to be the most teachable thing there is.

So let me put the question to you directly, the same one I put to myself with my hand on that tab. The next time you reach for the machine, are you going to think first? Or are you going to let it think for you?

I write about building with AI in the real world at BenSAIBrain.com. If you want help teaching your team to use AI as a thinking partner instead of a crutch, that is exactly the kind of work we do at Digital Ignitor.