The Day I Forgot How to Think

Last week, someone asked me a simple technical question. Something I would have known off the top of my head two years ago. And my first reflex wasn’t to think. It was to open Claude.

Not because the answer was hard. Because my brain, conditioned by months of cognitive delegation, decided it was easier to ask the AI than to access my own memory. And when I realized what I was doing — when I caught myself typing something I knew — I felt a discomfort I can’t ignore.

It’s not the first time. I’ve noticed my concentration on long texts has decreased. That I turn to Claude to draft things I used to write unaided. That I verify less because “the AI probably got it right.” Each of these micro-decisions is insignificant alone. Together, they form a pattern that science now has a name for: Cognitive Offloading.

And the data I found about it genuinely alarmed me.

The Study That Should Be a Headline

Researcher Michael Gerlich (2025) conducted a mixed-methods study with 666 participants across diverse age groups and educational backgrounds. The methodology combined surveys with in-depth interviews, using standardized critical thinking measures — both self-assessment and performance-based evaluations.

The key finding: strong negative correlation between frequent AI tool usage and critical thinking abilities, mediated by cognitive offloading.

In plain language: the more you use AI indiscriminately, the worse your critical thinking gets. And the mechanism isn’t mysterious — it’s biological. If you don’t exercise a muscle, it atrophies. If you delegate all your intellectual decisions to an algorithm, your ability to perform them independently degrades.

The study identified what Gerlich called “cognitive laziness” — a decline in the inclination to engage in deep, reflective thinking, as a consequence of persistent AI reliance. It’s not that people can’t think anymore. It’s that they don’t want to — because the reflex of “ask the AI” has become more automatic than the reflex of “think about it.”

Kim et al. (2026), in a conceptual review integrating multiple empirical streams, document that the fluency with which AI provides solutions creates a feedback loop where users progressively delegate more cognitive work, with the long-term consequence of atrophying their own capacities.

The Data Point That Chilled Me

The International AI Safety Report 2026 — the most comprehensive official document on AI risks published to date — included cognitive offloading as a formal risk. And cited a particularly alarming medical study:

Three months after introducing AI support, clinicians’ ability to detect tumors without AI assistance dropped by 6%.

Six percent. In three months. Doctors who knew how to diagnose were losing the ability to do it without the tool. Not because they became worse doctors — because their brains adapted to the support’s presence and reduced cognitive investment in the task.

A March 2026 study in Frontiers in Psychology, with university students, confirmed the pattern: cognitive offloading through digital tools is negatively associated with critical thinking, task persistence, and learning depth.

Young people (17-25) showed higher AI usage, greater cognitive offloading, and lower critical thinking scores compared to older participants. The most “digitally native” generation is the most vulnerable — not despite their familiarity with technology, but because of it.

We’ve Seen This Movie Before

Cognitive science recognizes that cognitive offloading isn’t new. We’ve done it before:

GPS replaced our ability to memorize routes and read maps. Research shows people who constantly use GPS have worse spatial orientation than those who navigate unassisted.

Calculators replaced the need for mental arithmetic. Few adults today can do long division manually — a skill that was routine 30 years ago.

Smartphones replaced our memory for phone numbers, dates, and facts. The “Google effect” — the tendency to forget information we know is available online — has been documented since 2011.

But AI is operating this offloading at a completely different level. It doesn’t just replace physical effort or basic math. It replaces the ability to think, create, and solve problems independently. It’s the first technology that directly attacks higher-order cognitive functions — exactly the ones that make us human.

The Business Strategy: A Customer for Life

Here’s the part that makes me most uncomfortable. The world’s most powerful AI tools are free. Maintaining this infrastructure costs fortunes — data centers, energy, NVIDIA chips. The math doesn’t work.

Unless the business model isn’t selling the product. It’s creating dependency.

The sequence is predictable: the bait (free model attracts the user), the habit (the tool becomes indispensable in the daily workflow), the dependency (the user loses confidence or ability to execute without AI), and the monetization (lifetime customer locked into a subscription because they can no longer function without it).

Sam Altman has publicly said he sees digital intelligence as a public utility — like electricity or running water. People will pay through a “meter.” And when that meter turns on, who will you be without it?

As the Kim et al. (2026) paper describes: the fluency with which AI provides solutions creates a feedback loop where we progressively delegate more and atrophy the ability to do it ourselves. It’s not conspiracy. It’s design.

Tool vs. Crutch: The Line I Cross Every Week

This is the distinction I struggle to maintain — and frequently fail at:

AI as tool: using it to accelerate something you already know how to do. Critically validating every response. Using saved time to think about strategy — not to produce more.

AI as crutch: letting AI do tasks because you’re too lazy to learn. Accepting the first response with blind trust. Using saved time to produce more mediocre content instead of thinking better.

In practice, the line is blurred. I catch myself on the “crutch” side more often than I’d like to admit. And each time I notice, I implement a deliberate “do without AI” exercise — write a complex email unaided, solve a technical problem in my head, read a long article without asking for a summary.

Not because I’m a masochist. Because I know my brain needs the exercise to avoid atrophying. It’s the cognitive version of going to the gym.

What Research Suggests as Antidote

Gerlich’s paper and Kim et al.’s review offer directions that work:

Structured prompting. Instead of asking AI for the complete answer, ask it to pose questions that guide you to the answer. “Help me think about this problem” is cognitively healthier than “give me the answer.”

Forced justification. Before accepting an AI output, force yourself to explain why the answer is correct. If you can’t explain it, you didn’t understand it — and you’re delegating, not learning.

Intermittent retrieval practice. Periodically, do tasks without AI. Write, calculate, analyze in your head. Keeping these circuits active is what prevents atrophy.

Moderate use is positive. The IE University paper (October 2025) is emphatic: moderate AI use can have positive cognitive impact. The problem isn’t using. It’s using indiscriminately, without awareness of the effects.

Conclusion: Who Will You Be When the Meter Turns On?

AI is here to stay. Ignoring it would be a strategic mistake. But the real secret to professional success in the coming years is ensuring technology is your lever, not your crutch.

If tomorrow all these tools were placed behind an expensive paywall and you couldn’t afford them, what would remain? Just your brain. And you’d better make sure you still know exactly how to use it.

I ask myself this every week. I don’t always like the answer. But the fact that I’m asking is already half the defense.

Share if this resonated:

If the product is free, the product is you. But with AI, it’s not your data being collected. It’s your ability to think being switched off.

Read Also