There is a long tradition of worrying that new tools will make people dumber. Socrates worried about writing. Educators worried about calculators. Parents worried about search engines. In each case, the fear was overstated, and the tools changed how people think in ways both good and bad.
AI assistants are the latest chapter, and the stakes are higher because they sit closer to the act of thinking itself.
The outsourcing temptation
The most obvious effect of AI on thinking is the temptation to outsource it.
Why compose a careful email when the assistant can draft it? Why work through a problem when a summary is a keystroke away? Why remember anything when the machine can retrieve everything? Each of these is a real convenience, and each, taken to its extreme, is a form of mental atrophy.
The danger is not the tool. It is the unexamined habit: reaching for the assistant first, before the mind has had a chance to engage. The question is not whether AI helps thinking. It is whether people use it to replace thinking or to amplify it.
The good version: thinking with a partner
Used well, an AI assistant is less a substitute for thought than a partner in it.
A good assistant can do the legwork — retrieving, summarizing, organizing — so the human can do the work that matters: questioning, evaluating, deciding. It can surface alternatives you had not considered and stress-test ideas you were too attached to. It can be a sounding board that never gets tired.
This is the productive pattern. The machine handles the processing; the human handles the judgment. The outcome is better thinking, not less of it — provided the human stays in charge of the conclusions.
The hidden cost: fluency without understanding
The subtle risk of AI-assisted thinking is fluency without understanding.
An assistant can produce a polished summary of a topic in seconds. The summary is easy to read, easy to absorb, easy to believe. But the reader who has not wrestled with the underlying material has not built the framework to evaluate it — to spot what is missing, to notice the weak argument, to connect it to what they already know.
Fluency is seductive because it feels like mastery. The danger is that people mistake the smoothness of the output for the depth of their own understanding. The machine understood it; the human only recognizes it.
What actually changes in the brain
Cognitive scientists describe thinking as partly a process of “offloading” — using the world to store and process information.
A notebook offloads memory. A spreadsheet offloads calculation. AI offloads a much broader category: synthesis, comparison, even argument. Offloading is not inherently bad; it is how human intelligence has always worked. But what you offload, you do not practice — and what you do not practice, you may lose the facility for.
This is the real trade. The skills you hand to the machine are skills you stop exercising. Whether that matters depends on which skills you hand over, and whether you can get them back when the machine is not around.
The writing case study
Writing is the clearest example, because writing is not just a way of recording thought — it is a way of thinking.
When people write by hand, they organize, clarify and deepen their thinking in the process. When an AI drafts a text and the human merely edits, that organizing work is skipped. The result is often a better text and a shallower engagement with the subject.
This does not mean drafting with AI is wrong. It means the human should do the thinking first — outline, argue, struggle — and use the machine for what it is good at. The thinking that happens before the draft is the thinking that belongs to you.
The habits worth keeping
Given all this, there is a practical set of habits that protects thinking in an AI-assisted age.
First, think before you ask. Frame the question, form your own hypothesis, attempt your own answer — then consult the machine. Second, argue with the output. Do not treat the assistant’s answer as the answer; interrogate it, disagree with it, verify its claims. Third, keep a zone of unassisted practice — writing, calculating, reasoning — where you exercise the skills yourself.
None of this is anti-technology. It is pro-judgment. The machine is a remarkable partner; the question is whether it has a partner or a servant.
The honest conclusion
The machines that help us think will, without doubt, change how we think. That was true of writing, calculators and search. It is true of AI.
Whether the change is for the better is not predetermined. It depends on choices that are being made every day: whether people delegate the thinking or just the processing, whether they treat AI as an oracle or a colleague, whether they keep their own minds in the loop.
The question of who owns the thinking
Underneath the habits and the tools is a sharper question: who owns the thinking in an AI-assisted workflow?
If the machine drafts, summarizes and suggests, and the human approves, the output may be excellent while the intellectual ownership has quietly migrated. The work reflects the machine’s style, its blind spots, its preferences. The human has become a reviewer rather than an author.
This is not inherently wrong — but it is a choice, and it should be a conscious one. For work that matters, the thinking worth owning is the thinking you do before you delegate: the framing, the judgment, the stance. Protect that, and the machine becomes a tool. Hand it over, and you have a collaborator — whether you noticed the transfer or not.
The tool will not decide what thinking is worth. We will. And the people who use AI to think better, rather than instead of thinking, are the ones who will get the best of both — a faster mind and a deeper one.