In partnership with
Two weeks ago I was reading a paper on monetary policy and got to a sentence I didn’t fully understand. Before I had finished the sentence, my cursor was in the Claude tab.
I hadn’t tried to figure it out. I hadn’t even named what I didn’t understand. I had just reached.
I closed the tab and went back to the paper. The sentence took me about ninety seconds to parse. I felt the small discomfort of having to think-and the smaller satisfaction of having actually thought.
The reaching had become reflexive. Not the using. The reaching. The skip past the part where I sit with the difficulty for a beat.
Free Weekly AI Sessions for Experienced Software Engineers.
Every Wednesday at 5 PM CT, Gauntlet AI professors teach a live, hands-on AI engineering session — completely free. If you're nontechnical, this isn't for you. New topic every week, built for engineers who want to build, not just watch. See upcoming sessions.
What you stop doing
Most arguments about AI and thinking go abstract-civilization, education, the future of work. The useful conversation is small and behavioral: what do you stop doing in a typical hour, now that an AI is always available?
What you stop doing is the few seconds of effort it takes to hold an unresolved question in your head. The pause. The “wait, what does that actually mean?” before you ask anyone.
That pause is where most of what we call critical thinking lives-not in the eventual answer, but in the friction of formulating the question.
Cognitive offloading is the technical term. We’ve always done it-notebooks, calculators, Google. Each transferred mental work to a tool and freed us for what supposedly mattered more.
AI is different. It doesn’t just store information or run a computation. It produces reasoned-looking output. What it offloads isn’t memory or arithmetic-it’s the practice of working something out.
David, a product lead I worked with last year, ran into this. He’d always written his own briefs-long, messy first drafts he’d cut and refine.
After six months of AI drafting, his thinking on hard product questions had gotten shorter-not faster, shorter. He could no longer hold three competing framings in his head the way he used to.
He didn’t quit the tool-he changed the order. Ten minutes by hand first, ugly and unfinished, before the AI saw the problem. Then the AI challenged his thinking instead of replacing it.
The drafts got better. So did the thinking under them.
What the study found
In January 2025, Michael Gerlich at SBS Swiss Business School published a study of 666 UK participants across three age groups.
He measured AI use, cognitive offloading, and critical thinking with the Halpern Critical Thinking Assessment.
He found a significant negative correlation between frequent AI use and critical thinking scores, mediated by cognitive offloading. Participants aged 17–25 used AI the most and scored lowest. Higher education attenuated the effect-but didn’t eliminate it.
One caveat worth holding: this is correlation, not proof of cause. Self-reported use, snapshot in time.
The study doesn’t prove AI is making you worse. It shows the pattern is there, strongest in the youngest users, worth watching in yourself.
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
What to try this week
For the next week, try this. When you’d normally open the AI tab, set a two-minute timer and formulate the question first. Write it down in one sentence. Then decide if you still need the AI.
You’re not boycotting the tool. You’re keeping the small muscle that decides when the tool is the right answer.
The full paper is here.
— Prompt N Productive—