Is AI Making Us Dumber? The Science of Cognitive Offloading

Does AI make us dumber? Not in the way the scariest headlines imply — there's no evidence that using ChatGPT lowers your IQ or rewires your brain for the worse. But the research points to something more specific and more useful: when you let a tool do your thinking, the skill you handed over gets rusty from disuse. AI doesn't subtract intelligence. It just makes it very easy to stop practicing.

That distinction matters, because it changes the advice. If AI were simply harmful, the answer would be to avoid it. It isn't, and you shouldn't. The real question is whether you're using AI to extend your thinking or to replace it — and the science of cognitive offloading is the best lens we have for telling those two apart.

What cognitive offloading actually means

Cognitive offloading is the technical term for using an external aid to reduce the mental effort a task would otherwise require. Writing a phone number down instead of memorizing it is offloading. So is a calculator, a GPS, a sticky note — and now, a large language model that drafts your email or reasons through a problem for you.

Offloading is usually rational. Your working memory is limited and your attention is finite, so handing routine work to a tool frees you up for harder things. The catch is that your brain optimizes for whatever you actually do. If a reliable external system is always available, the brain quietly stops doing the work itself.

Offloading isn't cheating. It's efficient. The problem is only that efficiency, repeated daily, slowly erodes the skill you stopped using.

The "Google effect": the warning shot from 2011

The clearest early demonstration came more than a decade before generative AI. In a 2011 study published in Science, researchers Betsy Sparrow, Jenny Liu and Daniel Wegner ran a series of experiments on what they called the "Google effect." They found that 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.

In other words, the brain treated the search engine as an external hard drive and reallocated effort accordingly. That isn't necessarily bad: knowing where to find an answer is itself a skill. But it confirmed a principle that AI now stretches much further — when an external memory is dependable and always on, we offload to it automatically, often without noticing.

Generative AI raises the stakes, because it doesn't just store facts the way a search engine does. It performs the reasoning, drafts the prose, and reaches the conclusion. The thing being offloaded is no longer just memory — it's judgment.

What the 2025 research is finding

Several studies published in 2025 have begun to map what happens when we offload reasoning, not just recall. Read carefully, they tell a consistent and fairly measured story.

A few things are worth saying plainly about this evidence.

First, these are correlations, not proof that AI causes a decline. It's entirely possible that people who already lean less on critical thinking are also drawn to lean more on AI. The Gerlich and Lee studies rely substantially on self-report, which has its own limits.

Second, the most dramatic of the three — the MIT Media Lab's "Your Brain on ChatGPT" — should be read with real caution. It's a small study (54 participants), it has not been peer-reviewed, and other researchers have publicly flagged concerns about its EEG methods and sample size. It's suggestive, not conclusive, and it would be a mistake to treat "lower brain connectivity" as proof of damage. It points in the same direction as the survey work, which is the most you can honestly say.

What ties all three together isn't "AI rots your brain." It's that the mental moves you delegate are the ones you stop rehearsing — and the Lee study adds a crucial nuance: the people most at risk are the ones who trust the tool most, because they shift from doing the work to merely checking it, and sometimes skip the checking too.

The skills most exposed to offloading

Not everything offloads equally. The abilities most at risk are the ones AI is genuinely good at imitating:

• Critical thinking. AI is fluent and confident, and sometimes confidently wrong. If you accept its output without interrogating it, you stop practicing the judgment that catches the error. (Our guide on how to improve critical thinking goes deep on this.)
• Memory and recall. Why retain a fact you can re-query in seconds? Often there's no reason — except that recall is also how you connect ideas and notice when something doesn't add up.
• Composition and articulation. When a polished paragraph is one prompt away, the muscle that turns a vague thought into a clear sentence gets little use. That's a real loss, because communication is becoming a rarer and more valuable edge, not a less important one.

How to use AI without going rusty

The goal isn't to use AI less. It's to stay sharp enough to use it well. A few deliberate habits make most of the difference.

1. Do the rep first, then reach for the tool. Draft the hard sentence yourself before asking AI to improve it. Estimate the number before you compute it. Try to recall before you search. You can still use AI to go further — you just don't skip the rung.
2. Treat AI output as a first draft, not a verdict. Ask "how would I know if this were wrong?" The most dangerous moment is when an answer is fluent and you're tired.
3. Match your trust to your own expertise, not the tool's confidence. The Lee study's lesson is that misplaced confidence in AI is exactly where critical thinking quietly disappears.
4. Keep a few skills off the AI entirely. Mental math, on-your-feet speaking, reading a long argument closely — protect a handful of abilities you train without assistance, the way you'd protect a muscle you don't want to atrophy.

Key takeaways

• AI doesn't make you inherently dumber; disuse does. The skill you offload is the skill you stop practicing.
• The 2011 "Google effect" showed we offload memory automatically; 2025 research suggests we now offload reasoning too.
• The strongest evidence is correlational, and the most striking study (MIT's EEG preprint) is small and not peer-reviewed — read it as a signal, not a verdict.
• The fix isn't avoidance. It's deliberate use: do the rep first, verify the output, and keep a few skills sharp on your own.

This is the philosophy behind Rusty — a brain gym for staying sharp in exactly this environment. Each five-minute circuit gives you reps in the skills AI is happy to take from you, and it flips the script by putting AI on your side: answer a speaking prompt out loud and you get an instant Clarity Score on your structure, reasoning and delivery, instead of letting a model do the talking for you.

The most valuable thing you can do as AI gets better at thinking is refuse to let that be the only thinking that happens.

Want to keep your edge? Download Rusty free on the App Store and run your first circuit in five minutes.