Is AI Rotting Our Brains?

 

It's a question all of us have on our minds as we rely more and more on machines throughout our workflows and, for many of us, in our personal lives. Everywhere we see AI transformation, the question mostly heard is: If I lean in and let AI make me a 10x person, what will happen to my brain?

My answer is simple, if not vague: it depends. That’s what I’ll unpack in this long-form article. Keep reading until the end, when I share four rules for getting the most out of AI while keeping yourself sharp as a tack throughout adoption.

Humans have worried about cognitive offloading for a long time. In his influential 2016 study on the subject, Sam Gilbert defined cognitive offloading as using physical action to change the information-processing demands of a task, thereby reducing the mental effort required. For example, when you tilt your head to view a rotated image, you physically change the task so your brain no longer has to mentally rotate it. That’s cognitive offloading. The simplest definition of cognitive offloading is taking information processing out of your head and handing it off to something outside your body.

As Gilbert puts it, humans have always been “mental misers,” conserving the effort we ask our brains to expend. We’ve used technology since the dawn of civilization to save mental effort, going all the way back to the invention of the written word. And that’s when the first fears began.

In ancient Greece, Socrates was famously skeptical of the written word and the effect it would have on our minds. Before widespread writing, knowledge depended far more on memory and was passed by word of mouth from one person to another, requiring people to hold far more information in their heads for recall.

In Plato’s dialogue Phaedrus, Socrates recounted a myth about the Egyptian god Theuth presenting writing to King Thamus as a tool that would make people wiser and improve their memories. Thamus rejected that promise, warning that writing would “produce forgetfulness” and give people only the appearance of wisdom.

But over the centuries, the written word and later the printing press helped drive enormous advances in society and in our capacity for critical thinking.

Fast forward to the mid-1970s, when pocket calculators entered the mainstream as affordable devices that enabled students to offload basic calculations to a handheld device.

Critics and many educators argued that letting students use pocket calculators would destroy their ability to perform basic mental arithmetic, make children lazy, leave them dependent on batteries, and prevent them from detecting simple mathematical errors without a screen supplying the answer. Some teachers even protested their use.

Years later, a broad body of research found that calculators could free students to learn more complex mathematics rather than weaken their abilities. In turn, that helped enable even greater advances in technology and human potential.

Then, at the turn of the 21st century, search engines transformed how we retrieved information, and the fear resurfaced. Studies warned that we would soon develop “Google brain,” with our growing reliance on search engines eroding our memory. We don’t have to remember anything anymore. We can just look it up later.

To date, there is little, if any, scientific research showing that our capacity for memory has declined. In fact, a study argues that we’ve developed a new cognitive skill, "meta memory": Knowing how to retrieve information through queries and, most importantly, remembering where that information can be found.

Then came GPS and navigation tools like Google Maps. Instead of learning how to navigate a city, we let the machine give us turn-by-turn directions. Some still worry that this makes us overly dependent on having our phones, and cell service, to find our way around our own town.

Although books, calculators, search engines, and GPS haven’t led human civilization into widespread cognitive decline, we are still worried about AI because this time it’s different.

Tools like ChatGPT don’t just perform rote calculations, provide directions to a restaurant, or look up the schedule for your favorite football team. They can write entire essays on your behalf, create marketing plans, answer complex test questions, spin up websites, and make purchase recommendations.

In other words, they can think on your behalf. And thinking is one of our most fundamental human capabilities: the ability to evaluate, reason, debate, and ultimately make judgments.

But What Does the Research Really Say About AI’s Impact on Our Brains?

Let’s start with Michael Gerlich’s research at SBS Swiss Business School. His study, “AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking,” surveyed 666 participants about their AI use, cognitive-offloading habits, and critical-thinking skills. He also conducted in-depth interviews with 50 of them to better understand the patterns behind the data.

"The findings revealed a significant negative correlation between frequent AI tool usage and critical thinking abilities, mediated by increased cognitive offloading." - Michael Gerlich, SBS

As participants’ reported AI usage increased, their ability to “analyse, evaluate, and synthesise information to make reasoned decisions” tended to be lower.

It’s no surprise that this got the attention of other academics and captured the imagination of journalists. It was more than an opinion piece; it used established measures and in-depth interviews. That’s why its findings made their way into op-eds and slides on the lecture circuit.

But there are two important things to note about this research. First, the negative relationship was most pronounced among younger participants, who may not yet have fully developed their critical-thinking skills. Among older participants, the relationship between AI use and lower critical-thinking scores was much weaker.

Second, this study was a snapshot in time, not a longitudinal study that followed people over several years. It cannot tell us whether the decline in critical-thinking scores was permanent, temporary, or reversible. For all we know, reducing one’s reliance on AI, taking an AI-free weekend, going on vacation, or attending a stimulating in-person conference might help restore those mental muscles. Nothing in the research suggested that the effect was irreversible.

Next, let’s dive into the MIT Media Lab study titled “Your Brain on ChatGPT.” Researchers assigned essay-writing tasks to three groups. The Brain-only group wrote without digital tools, the Search Engine group could use traditional web search, and the LLM group used ChatGPT.

All participants wore EEG headsets so researchers could measure electrical activity and connectivity patterns in their brains as they wrote. That’s what makes the study fascinating. It didn’t rely only on surveys or self-reported behavior.

The Brain-only group showed the strongest and most widely distributed brain connectivity during the task. The ChatGPT group showed the weakest, suggesting that their brains had shifted closer to cruise control. Beyond the EEG findings, many ChatGPT users struggled to accurately recall or quote what they had written afterward and reported less ownership of their essays.

"While generative AI tools spare the user mental effort in the short term, overreliance leads to the accumulation of cognitive debt—a measurable reduction in independent memory consolidation, critical engagement, and executive control." - Dr. Nataliya Kosmyna, MIT Media Lab

Another interesting qualitative finding was that evaluators described the ChatGPT group’s essays as “soulless” and lacking original thought. That’s not entirely surprising given how large language models operate. They predict the next token based on patterns learned from enormous amounts of existing content, which can cause unedited writing to gravitate toward familiar ideas and phrasing.

As a result, an essay written almost entirely by AI can easily sound pretty average. (This article might have turned out that way if I had just one-shotted it through ChatGPT with a single prompt.)

While this study was praised as groundbreaking and widely quoted by researchers and journalists, it has its detractors. First, it included only 54 participants, a very small sample for conclusions this sweeping according to critics of the paper's methodology.

The paper also went viral as a preprint before undergoing peer review, and critics raised concerns about its study design, EEG methodology, reproducibility, and the generalization of its findings. Some worried that overstating the results could cause people to reject a promising technology prematurely. In fact, the brain-only group that later used ChatGPT had the highest performance of all!

More recently, in 2026, Boston Consulting Group released research examining what happens when generative AI use spreads across an entire organization. The report was titled “When Everyone Uses AI, Companies Risk Losing Critical Skills.”

The researchers focused on the concept of de-skilling, the loss of human capabilities as work is handed to machines in pursuit of efficiency. They argue that “unlike previous tools, GenAI doesn’t merely support human thinking. Increasingly, it substitutes for it, with consequences for a specific set of human skills.”

They then examined what this could mean for companies that achieve widespread AI adoption. In a global study of 70 C-suite leaders and senior executives, half said they were already observing de-skilling in their organizations, while more than 60% believed it would pose a material threat within the next three to five years.

While this sounds alarming at first glance, it’s important to understand that this was a small, nonrepresentative survey of a handpicked group of executives. They were not selected randomly, so the findings cannot be generalized to the broader business population.

For all we know, some of their responses may reflect their attitudes toward AI as much as what they’re actually seeing in the quality of work or decision-making across their organizations. Nevertheless, it’s an interesting finding.

While all of these high-profile studies are far from perfect, they do raise valid concerns. As the old saying goes, “Where there is smoke, there is fire.” The fire isn’t AI the tool. It’s AI as the great enabler for handing over more and more of our thinking.

Collectively, the research suggests we should be concerned about what excessive cognitive offloading could do to the quality of our work, the quality of our workforce, and perhaps the quality of our thinking. But the real issue may not be whether we use AI. It may be whether we use it to extend our thinking or replace it.

Does AI rot our brain or not?

Here's the good news: The answer is no.

Let’s dispense with the concept of “brain rot.” As researcher Ayuliem points out:

"Neurons don't decay from disuse the way the word implies. There's no necrotic tissue, no irreversible cellular death. The internet picked "brain rot" because it sounds scary, not because it's accurate. What's actually happening, if it's happening at all, is deconditioning, closer to muscle atrophy than decomposition. And that matters enormously because atrophy is reversible."

Think of it this way: If you broke your arm and had to wear a cast, the muscles in that arm would weaken while you weren’t using them. But once the cast came off and you started physical therapy, those muscles could regain their strength. Your cognitive skills can work in a similar way. When you stop exercising them, they may weaken. When you start using them again, they can recover.

But using ChatGPT to think on your behalf is like putting your brain in a cast. At some point, you’re going to have to take it off and put your brain through its paces to regain your mental strength. (Keep reading, because I’ll give you a way to avoid that mental cast altogether.)

As Ayuliem also notes, the MIT Media Lab research suggests that we can lose a sense of cognitive ownership over our work. In other words, the line begins to blur between what we know and what AI knows. We can develop the false sense of competence that Socrates warned about with the advent of the written word.

Here’s my biggest concern about over-reliance on AI, especially when we use it to produce presentations, reports, or articles like this one: We may never go through the struggle of researching, synthesizing, reasoning, organizing, and actually crafting the work ourselves.

As a result, we can appear shallow or struggle to answer questions during what I call the “freestyle section” of most conversations, the point where we have to leave the script and think on our feet.

Imagine one-shotting a report with Claude after connecting it to a set of spreadsheets. You present it to an executive, complete with the nifty slide deck Claude created to accompany it.

After you walk through the slides, the executive peppers you with questions. That’s the freestyle section of business life. This is where it can all break down. You don’t truly own the content, and as a result, you can neither unpack it nor defend it.

It’s no surprise that a recent study from Duke’s Fuqua School of Business found that managers and coworkers often judge people who use generative AI as less competent and lazier. That may not always be fair. But when someone can’t explain, unpack, or defend the work they’ve produced, the judgment becomes easier to understand. This is the risk of creating “work slop,” something I’ll write more about later.

So let’s dispense with the phrase “brain rot” and ask a better question: Is AI reducing our critical thinking skills? And, as I said at the top of this article, the answer is: It depends.

It depends on how much of your thinking you hand over to AI and how often you ask it to reason, evaluate, and generate ideas on your behalf.

It also depends on where you are in your development. The research repeatedly suggests that people earlier in their developmental journey may have more to lose when AI replaces the struggle required to build knowledge and judgment. All of this leads to a strategy that may help you enjoy the best of both worlds.

How Can You Lean Into AI’s Promise and Keep Your Brain Sharp?

You can get the most from AI while keeping your critical thinking sharp. But it requires intentionality. Here are four rules that can help.

Rule One: Give AI the Grunt Work, Not the Thinking Work

Here’s an example: Formatting a large spreadsheet, organizing the data into columns, or generating boilerplate code is grunt work. Let AI do it. But deciding which key performance indicators the spreadsheet should track or determining the underlying architecture of the code is thinking work. Keep ownership of that work yourself, even if you use AI to challenge or pressure-test your ideas.

Take how I wrote this article, for example. I used AI to conduct the first-pass research, finding relevant academic studies, critiques of those studies, and other media coverage.

Then I used an AI agent to fact-check the key research I was considering for the article. I loaded everything I had collected into NotebookLM, which turned the material into a podcast that I listened to while working out at the gym.

When I got home, I stepped away from AI, took out a pen and paper, created an outline, and wrote the article myself. That’s not to say I didn’t use AI during the writing process. I just used it in a different way than you might expect.

I used a generative AI voice keyboard called Willow Voice to do the typing. I press the function key, speak, and it types for me, about three times faster than I can type on my own. It increases my productivity while preserving my natural voice.

During the editing process, I used ChatGPT to suggest places where I could tighten the wording or improve clarity. But I didn’t let it rewrite the work.

Through this process, I got the best of both worlds. I eliminated much of the grunt work, but the work itself was still a valuable struggle.

Rule Two: Create and Follow Your Developmental Roadmap

Ask yourself: What skills do you want to develop for yourself as part of your career? Which strengths do you want to keep building, and which weaknesses simply aren’t worth mastering?

In my case, writing is an ongoing skill I need to continue to improve. Information retrieval is not a capability I need to build deeply, so I gave that job to AI.

The same goes for image design, including the cover of this newsletter, and the complex math behind expected-value calculations for the bets I make with my time and money. Both capabilities offer me tactical value, but I don’t have the time to master Adobe Illustrator or trigonometry while keeping up with the demands of my day job. They’re not on my developmental roadmap, so I let AI handle them. That’s how I think about the division of labor.

Rule Three: Build Your AI Management Muscle

Think about how you write prompts. Don’t just give AI an assignment, hit return, and hope for the best. Be more strategic. When you ask ChatGPT to generate insights, summarize a complicated subject, or work with you on a project, first break the assignment into the steps it should take. This is called task decomposition, and it requires you to think carefully about what a good reasoning process should look like. That, in itself, is a form of metacognition.

Next, give ChatGPT a rubric for what success looks like. Better yet, find examples of excellent work and provide them as models to follow. Then review what it gives you and compare it against those standards. When you question the result, ask it to explain its rationale, assumptions, and sources.

But most importantly, own your work. Always review what AI gives you for quality, consistency, and fit with how your organization operates. Slop is a decision, not the result of Critique it and iterate. Slop is a decision!

These steps will not only improve the quality of what you get back from AI; they’ll also help you develop a new kind of productive struggle: becoming an agent boss.

Speaking of which, if you haven’t done so already, build and deploy your first agent as soon as you can. Give it one recurring task in your weekly workflow, especially if you already have access to tools like Claude Cowork. You can create skill files for it or simply start with one small task you regularly perform.

This will be a struggle because it’s not just about giving an agent instructions. You’ll also have to manage it, evaluate its work, and help it improve through ongoing feedback. Taking on this new challenge of building your AI management muscle can replace some of the productive struggle you’ve surrendered in the name of efficiency.

And believe you me, this belongs on your developmental roadmap. Learning to manage AI agents will be a critical skill for just about everyone.

Rule Four: Keep Your Brain Challenged

Remember, it’s the struggle of learning and improving that helps keep your cognitive abilities strong. At work, block off two one-hour sessions each week to tackle hard problems without using AI. Just you, perhaps collaborating with other people or working something out on a whiteboard, but without handing the thinking over to a machine. That’s only two hours out of a 40-hour week, but it can go a long way toward keeping your brain sharp.

It’s also helpful to develop hobbies or learn new skills that feed your curiosity or complement your developmental roadmap. Some people learn a new language because they plan to visit the country. Others do it simply because the language fascinates them. Research suggests that multilingualism can strengthen connectivity between regions of the brain and may help slow cognitive aging.

For me, I pour myself into learning modular synthesis, which allows me to make music without using a computer. Here’s an example of the ’70s throwback synthesizer rig I built. It’s about as anti-AI as it gets.

Do better things

It’s important to look at AI not through the lens of “How much time can I save?” but through the lens of “How much greater can my work become?” Christopher Dede, a senior research fellow at Harvard, puts it this way:

“Ninety-five percent of what I read about AI in education is that it can help us do things better, but we should also be doing better things. One of the traps of GenAI, even when you’re using it well, is that if you’re just using it to do the same old stuff better and quicker, you have a faster way of doing the wrong thing.”

I think he makes a really good point. For me, the goal is simple: I want to elevate the value I create in the world. If AI can help me do that, that’s great.

A few weeks ago, I was giving a lecture on AI transformation when one of the attendees asked me, “As you use AI in your content process, are you saving a lot of time? Do you write these LinkedIn articles faster than you did before?”

And I told him, “No. I’m not writing these pieces faster. I’m writing better pieces: longer, richer, more useful to the reader, and more likely to perform well in AI search. Because AI has taken away much of the grunt work, I have more time to think deeply and create something genuinely valuable. And ultimately, the internet rewards content that gives people real value.”

So that’s my final word to you: AI is a superpower. Use it not just to become a more efficient worker, but to become a more capable, creative, and valuable human being.