A study I read recently has stuck with me. It found that the people who lean hardest on AI tools tend to score lowest on critical thinking, with the steepest gap among the young.
I use AI tools every day for parts of my own writing work, so this is not a doom blog about technology rotting our brains. It is more that I read the study, recognized something in my own habits, and wanted to think through what it actually says before deciding how worried to be.
I am not a psychologist or a cognitive scientist, and I am not credentialed in any of this. What follows is one writer reading the research and thinking aloud about it. The study at the center of this is observational, and a population-level pattern is not a verdict on you or on anyone in particular.
What the study actually found
The work comes from Michael Gerlich at SBS Swiss Business School, published in early 2025 in the journal Societies. He used a mixed-method design: a survey of 666 participants across a range of ages and education levels, plus 50 follow-up interviews to put some texture on the numbers.
The headline result is blunt. As Gerlich put it, “The findings reveal a strong negative correlation between frequent AI tool usage and critical thinking abilities, mediated by cognitive offloading.”
Before that sentence does too much work, the word that matters most is correlation. Gerlich is careful about this, and so should we be. He notes that “the study highlights correlations rather than direct causation, meaning that while AI usage and cognitive offloading are linked to reduced critical thinking, additional factors may also contribute to this relationship.”
The reverse direction is also possible: that people who already think less critically may simply reach for the tool more. The study cannot tell you which way the arrow points.
The cognitive offloading mechanism
Cognitive offloading is the habit of delegating a mental task to something outside your head, a calculator, a sticky note, a search bar, an AI. We have always done it. The question the study raises is what happens when the thing we offload to is good enough to do the whole job, not just hold the answer.
There is an experimental companion piece worth knowing about, though with heavier caveats. An MIT Media Lab study by Nataliya Kosmyna and colleagues had 54 people write essays in three conditions, using an AI assistant, a search engine, or nothing but their own heads, while measuring brain activity. The brain-only writers showed the strongest neural connectivity; the AI users the weakest, and they struggled to quote work they had just produced. The main sample is small, the follow-on crossover phase included only 18 participants. Treat it as a suggestive clue rather than proof. Even with those caveats, it points at the same intuition from a different angle: when the tool ends up doing most of the assembling, less of the assembling is being done by the writer.
The young show the sharpest effect
The age pattern is the part of the study that is most worrying. The youngest band, aged 17 to 25, used AI most and scored lowest on critical thinking, while those 46 and above used it least and scored highest. Higher education was protective across the board.
Gerlich’s reading of this is appropriately tentative, and I want to keep it that way. He suggests that “digital natives, who have grown up with AI-integrated technologies, might be more prone to cognitive offloading than older generations.” Note the hedges in his own sentence. The reading is softer than “young people are bad at thinking” — it is a researcher saying that people who never knew a world without these tools may lean on them more reflexively, simply because the tools were always there.
What to do with this
I don’t think the answer is to throw the tools away.
Gerlich himself lands on a balanced note: “AI tools are not inherently detrimental; rather, their impact depends on how they are used.” His own prescription, framed as his view rather than settled fact, is that “AI should complement cognitive engagement rather than replace it.”
That distinction is an important one. Whether AI extends my thinking or stands in for it has turned out to be the question that matters most in my own practice.
What I take from this is small and practical: notice which part of the task you are handing over. Outsourcing a lookup matters less than outsourcing the judgment about what the lookup means, and the second one is what I try to keep watch on.
None of this is a substitute for talking to someone qualified — a therapist or career counsellor, or a trusted person in your life — if the bigger questions here, about work, change, or where you stand, are weighing on you more than they are interesting you.
