To continue providing free, value-first guides and curated resources, some of the links on this site are affiliate links. If you click through and make a purchase, we may earn a small commission at absolutely no extra cost to you, which helps support the platform.
The AI Trap: Why Smart People Are Using It Wrong
Getting more likes than ever. Reaching more people than ever. And slowly disappearing as a thinker. This is what AI misuse actually looks like — and it is already everywhere.
- The friend who vanished
- Why smart people fall for it: the short-term trap
- The m-dash signal: how to spot AI-deferred thinking
- What the science shows: the MIT memory study
- What is actually happening in the brain
- The long-term risk nobody is talking about
- The one distinction that changes everything
- Are you using it wrong? A honest self-check
- Conclusion: a relationship, not a shortcut
The Friend Who Vanished
Think about someone you have known for years. You know how they write. You know their rhythms, their opinions, the particular way they construct an argument or land a joke. You know their voice.
Now imagine going to their LinkedIn profile one day and finding that voice gone. In its place: polished daily essays on leadership, productivity, and professional growth. Beautifully structured. Impeccably worded. Getting more engagement than anything they ever wrote before.
But it is not them.
This is not a hypothetical. Neuroscientist and brain health expert Dr. Daniel Amen described exactly this experience in a recent conversation about AI and the brain. He had a friend — someone he had known for a decade — whose LinkedIn profile had been transformed. Every day, a new essay. Every essay, reaching more people than the last. The friend was getting more likes and more reach than he had ever achieved writing in his own voice.
“Why would he go back?” Amen asked. And that question is the heart of the problem.
The friend was not being lazy. He was being rational — in the short term. He had found a tool that produced better-performing content faster than he could produce it himself. By every visible metric, it was working. And yet something important was being lost, quietly and invisibly, with every essay he did not write himself.
This is what AI misuse looks like in practice. Not a dramatic collapse. Not obvious incompetence. A gradual, comfortable, rewarded surrender of the thinking that makes you you — traded for content that performs better and costs you nothing. Except, eventually, everything.
Why Smart People Fall for It: The Short-Term Trap
To understand why intelligent, capable people misuse AI, you have to understand something fundamental about how human beings make decisions. We are not, as a species, particularly good at trading present comfort for future benefit. We know this. The evidence is everywhere.
The United States has a 75% obesity rate. Survey those people and ask whether they know that the cheeseburger is worse for them than the broccoli — the vast majority will say yes. Ask social media users whether they know their usage is making them more anxious — most will say yes. Then watch what they order. Watch what they reach for next.
Knowledge of the long-term cost does not reliably override the short-term reward. This is not ignorance. It is the basic architecture of human motivation, operating exactly as designed. We are wired to respond to immediate feedback, immediate reward, immediate relief. The consequences that arrive years from now are abstract. The dopamine hit that arrives in seconds is real.
AI fits this pattern precisely. Sitting down to think through a hard problem takes time, produces discomfort, and delivers uncertain results. Typing a prompt into ChatGPT takes seconds, produces polished output, and delivers immediate relief from the discomfort of not-yet-knowing. If that output also gets you more likes and more reach than your own thinking ever did, the reinforcement loop is complete. The behavior will repeat.
This is not a character failure. It is a predictable response to a system optimized to produce exactly this outcome. Understanding it as such is the first step toward choosing differently.
AI companies benefit when you use their tools more. Social platforms reward polished, high-volume content regardless of who or what produced it. Your audience cannot tell the difference and often prefers the AI version. Every signal in your environment is pointing toward more AI use and less of your own thinking. Swimming against that current requires understanding what is at stake.
What is at stake is not just authenticity or voice — though those matter. What is at stake, according to the neuroscience, is the health and capacity of your brain itself. The friend posting AI essays every day is not just losing his voice. He is losing the cognitive workout that writing in his own voice used to provide. And that loss compounds, invisibly, over time.
The M-Dash Signal: How to Spot AI-Deferred Thinking
There is a small typographical tell that has spread across the internet over the past two years so rapidly and so consistently that it has become, for those who notice it, an almost infallible signal of AI-generated text.
The em dash — this punctuation mark — was, until recently, relatively rare in everyday writing. Most people never used it. Style guides debated it. Many writers had never consciously deployed one in their lives.
Then ChatGPT arrived. And the em dash is now everywhere.
This is not because millions of people suddenly developed a fondness for this particular punctuation mark. It is because the large language models that power AI writing tools use em dashes heavily, and the people copy-pasting AI output into their posts, articles, and emails are importing the model’s stylistic fingerprints along with its words.
The em dash is the most visible marker, but there are others. Certain transitional phrases. A particular kind of listicle structure. The tendency to open paragraphs with a single bolded sentence followed by three sentences of explanation. The way complex ideas are broken into neat threes. A relentless, slightly inhuman evenness of tone. Once you know what to look for, you begin to see it constantly — in LinkedIn posts, in newsletters, in articles from publications that once had distinctive editorial voices.
In one well-known example, a group of friends confronted one of their own about an essay he had posted on LinkedIn — an essay that sounded nothing like him. They asked to see the prompt he had used. It was half a sentence long. “Write something about X issue.” That half-sentence had produced a two-to-three-page article he published under his own name. The prompt was the entirety of his intellectual contribution to a piece that went out to thousands of people as his thinking.
The problem with this is not plagiarism in the legal sense. The problem is what it represents: a complete outsourcing of the cognitive work that writing is supposed to do. Writing is not just a way of communicating ideas you already have. It is one of the primary ways you develop ideas in the first place. The act of writing — of finding the right word, constructing the argument, deciding what comes next — is itself a thinking process. When you replace that process with a prompt and a paste, you are not just changing your output. You are skipping the cognitive workout that would have made you sharper.
The em dash in someone else’s post is interesting. The em dash in your own post is a question worth sitting with: did you think this, or did you generate it?
What the Science Shows: The MIT Memory Study
The anecdotal evidence — the vanishing friend, the half-sentence prompt, the em dash epidemic — is striking. But the scientific evidence is more alarming still.
A study conducted at MIT examined what happens in the brain when people write using AI tools compared to writing without them. Participants were divided into two groups: one group wrote essays using ChatGPT, the other wrote without any AI assistance. Researchers then measured brain activity during the writing process and tested participants’ memory of what they had produced.
The results were stark.
Read that second figure again. Eighty-three percent of people who wrote an essay using ChatGPT could not accurately recall what their essay said — minutes after finishing it. They had produced content they did not own, in any meaningful cognitive sense. The words existed on the screen. The ideas had not passed through their minds in any way that left a trace.
This is not a flaw in the tool. It is a predictable consequence of how memory works. Memory is not a recording device. It is an encoding process, and encoding requires experience. You have to do something with information — wrestle with it, connect it to what you already know, make decisions about it — for it to be transferred from working memory into long-term storage. Passive exposure does not encode. Effortful processing does.
When you ask AI to write an essay and then read the result, you are a passive reader of someone else’s process. The thinking that would have encoded the ideas — the friction of finding the right phrase, the decision about what to include and what to leave out, the moment of genuine surprise when an argument takes an unexpected turn — none of that happened for you. And so none of it was encoded.
The MIT study showed that AI-assisted writing produced nearly two times less activity in the brain region linked to memory. If you are using AI to produce content you intend to know, understand, or build on later — reports, proposals, learning notes, professional writing — passive AI generation may be actively undermining your ability to retain what you produce.
What Is Actually Happening in the Brain
To understand why the MIT results look the way they do, you need to understand a basic principle of neuroscience: the brain is a use-it-or-lose-it organ. Neural pathways that are regularly activated become stronger, faster, and more efficient. Pathways that are not used weaken and, over time, atrophy.
This is not metaphor. It is the biology of the brain. Myelin — the fatty sheath that insulates neural connections and dramatically increases their transmission speed — builds up around pathways in proportion to how often they are used. When you practice a skill, think through a complex problem, write in your own voice, or struggle to find the right word, you are not just completing a task. You are physically strengthening the brain structures that make those activities possible.
When you outsource those activities to AI, the biological maintenance does not happen. The neural pathways for sustained reasoning, for creative synthesis, for the particular cognitive effort involved in putting your own ideas into words — these pathways receive no activation signal. They do not get stronger. They get weaker.
The hippocampus is particularly relevant here. This brain structure is central to memory encoding and to the kind of spatial and conceptual navigation that complex thinking requires. Research consistently shows that the hippocampus responds to challenge and novelty — it thrives on the cognitive effort involved in genuine problem-solving. When that effort is removed, hippocampal engagement drops. And hippocampal health, we now know, is one of the most significant predictors of cognitive aging and dementia risk.
The calculator analogy is instructive here — and genuinely hopeful. When electronic calculators were first introduced, there were serious concerns that they would atrophy mathematical ability. The long-term evidence suggests something more nuanced: calculators freed up cognitive space that humans redirected toward higher-order mathematical thinking. The tool reduced low-level computational load, allowing the brain to engage with more complex problems.
This is the version of AI use that protects and develops the brain: using AI to handle the mechanical or low-level so that your cognitive energy can go toward the higher-order. The version that damages it is using AI to handle the higher-order thinking itself, while you contribute only the prompt.
The Long-Term Risk Nobody Is Talking About
Dementia is not a switch that flips. It is a process that unfolds over decades, shaped by the accumulation of countless small decisions about how the brain is used. The concept of cognitive reserve — the brain’s resilience against neurological damage — describes the protective buffer built through a lifetime of mental engagement: education, complex work, social interaction, learning new skills, intellectual challenge.
People with high cognitive reserve show significantly later onset of dementia symptoms, even when the underlying biological changes associated with Alzheimer’s disease are present. The brain with high reserve has more pathways, more redundancy, more capacity to compensate for damage before it becomes symptomatic. The brain with low reserve has fewer options when damage begins.
Cognitive reserve is built by doing hard mental things. It is depleted — or rather, fails to accumulate — when mental effort is chronically avoided. And this is where the long-term picture of habitual AI misuse becomes genuinely concerning. If a significant portion of the population spends the next twenty years delegating the effortful parts of thinking to AI tools, the cumulative effect on cognitive reserve at a population level is not trivial.
Nobody is suggesting that using AI for a LinkedIn post causes dementia. But the habit pattern that produces the LinkedIn post — chronic avoidance of cognitive effort in favor of AI-generated output — is the same habit pattern that, extended across years and decades, systematically fails to build the cognitive reserve that protects against age-related cognitive decline.
We are in the early years of mass AI adoption. The long-term consequences for brain health will take a generation to become fully visible — perhaps twenty years or more. By the time the data is in, the habits will be deeply set. This is exactly the window in which the decisions individuals make about how they use AI matter most. Prevention is possible. Reversal, later, is harder.
The people most at risk are not those who use AI occasionally. They are those who use it habitually as a replacement for their own thinking — who, over years, build a relationship with AI in which the tool does the cognitive heavy lifting and they provide only the request. That relationship feels productive. It looks productive. Its costs will not show up on any metric for a long time. But they will show up.
The One Distinction That Changes Everything
Everything in this article converges on a single distinction. It is not complicated. But it is almost never made explicit, which is why so many intelligent people end up on the wrong side of it.
You provide a brief prompt. AI produces the output. You review it, possibly edit it lightly, and publish or submit it. Your cognitive contribution: minimal. Your learning: minimal. Your brain activation: low. Your memory of what was produced: poor.
You think first, draft first, or at minimum form a clear view. You use AI to challenge, extend, refine, or pressure-test your thinking. You ask it to explain, justify, push back. You write your own synthesis. Your cognitive contribution: central. Your learning: real. Your brain activation: high.
The difference between these two modes is not the tool. It is the role you assign yourself in relation to the tool. In the first mode, you are a passive recipient of AI output. In the second, you are an active thinker using AI as an unusually capable thinking partner.
Think of the difference between a student who copies answers from a textbook and a student who uses the textbook to check and extend answers they have already worked through. Both students have access to the same resource. Only one is learning. Only one is building the neural architecture that will serve them in ten years, twenty years, a lifetime.
The calculator freed up cognitive space by handling computation, allowing humans to engage in higher-level mathematics. The misuse of AI inverts this: it handles the higher-level thinking, while the human contributes only the low-level request. The tool that could make you a sharper thinker instead makes you a more efficient passive consumer of thought you did not have.
The question is not whether to use AI. The question is which side of that distinction you are on — and whether you have been honest with yourself about the answer.
Are You Using It Wrong? An Honest Self-Check
The trap is comfortable precisely because it looks like productivity. You are producing more. You are getting better results by visible metrics. The cost is invisible, deferred, and accumulating quietly. These questions are designed to bring it into view.
- When you read something you wrote with heavy AI assistance, can you reconstruct the reasoning behind it? Or do you find yourself re-reading it as though someone else wrote it — because someone else effectively did?
- Do you form your own position on a topic before consulting AI, or do you go to AI first and adopt whatever framing it provides?
- If you had to write your last three AI-assisted pieces without AI right now, could you produce something of equivalent quality? If the answer is no, the tool has been doing thinking you needed to do yourself.
- Has your tolerance for cognitive difficulty decreased? Does writing in your own voice feel harder, slower, or more frustrating than it used to? That frustration may be the atrophy talking.
- Would the people who know you best recognize your voice in what you are publishing? Or have you, like the LinkedIn friend, been replaced by a more productive, more polished, less you version of yourself?
- When AI produces something you did not expect, do you question it, push back, ask it to explain — or do you simply use it? The latter is passive consumption. The former is active thinking.
- Are you learning through your AI use — do you know more, think more clearly, reason more effectively because of how you use it? Or are you producing more while understanding less?
These are not comfortable questions. They are not meant to be. The purpose of asking them is not guilt but clarity — because clarity about where you currently are is the only foundation for choosing something different.
A Relationship, Not a Shortcut
The friend posting AI essays every day is not a villain. He is a person responding rationally to the incentives in front of him, in exactly the way that 75% of Americans respond to the incentives of the food environment. The problem is not his character. The problem is that the incentive structure is pointing him toward something that serves him brilliantly in the short term and costs him something important — quietly, invisibly, cumulatively — over time.
The costs are not hypothetical. The MIT research is real. The neuroscience of cognitive reserve is real. The connection between habitual cognitive avoidance and long-term brain health is real. These are not alarmist projections. They are the beginning of an evidence base that will only grow clearer over the next decade.
None of this means AI is the enemy. Used well, it can make you a sharper thinker, a more productive worker, a more effective learner. The calculator comparison is genuinely apt: a tool that, used intelligently, frees cognitive capacity for more demanding work rather than replacing the demand altogether.
But “used intelligently” requires a clear-eyed understanding of the distinction between using AI to do your work and interacting with AI to get better work. That distinction is easy to articulate and consistently difficult to maintain — because the wrong side of it is faster, more comfortable, and better rewarded by every external signal in your environment.
Maintaining it requires something that the short-term incentive structure never provides on its own: a genuine understanding of what is at stake. Your brain is not just how you do your job. It is how you think, how you connect with other people, how you navigate the world, how you remain yourself as the years pass. It is the asset that everything else depends on.
Treat it accordingly.
Post 2: Your Brain on AI — What the Neuroscience Actually Shows. We go deeper into the biology of what happens to your brain under different patterns of AI use — and what the research on cognitive reserve means for the decisions you make today.
0 Comments