Cognitive offloading: what to hand to your tools, and what to keep
- Offloading means using lists, calendars, search or AI to save mental effort, and most of it is sensible.
- Early research suggests handing AI the reasoning itself can cost you skills you were building.
- Rule of thumb: offload storage, keep the thinking.
Cognitive offloading is using something outside your head to save mental effort: a list, a calendar, a calculator, a search engine, an AI. Most of it is sensible. It starts to cost you when you hand over the thinking you were meant to be getting better at.
A shopping list is cognitive offloading. So is the reminder on your phone and the sticky note on your monitor. Evan Risko and Sam Gilbert, who wrote one of the standard reviews of the subject in 2016, describe it as using physical action to change what a task asks of your mind. One of their examples is tilting your head to read text that's at an angle. People have always done this. It's a fair part of why writing was invented.
Where AI changes things
Most offloading takes over storage or routine steps. AI can take over the reasoning itself, which is newer, and early research suggests it matters. In a survey of 666 people in the UK published in 2025, Michael Gerlich found that heavier use of AI tools went with lower critical thinking scores, and that offloading explained much of that link. It's a correlation, so it can't say which way round the effect runs.
Stronger evidence came from a classroom experiment with nearly a thousand secondary school maths students. Those who practised with a plain ChatGPT-style assistant scored higher during practice, then did worse than students who'd never had it once it was taken away. A version designed to give hints rather than answers mostly avoided the drop.
Improvement during maths practice with AI
Compared with students practising without AI. Once the AI was taken away, the plain-chatbot group scored 17% lower than students who never had it; the tutor version largely avoided that drop. Bastani et al. (2025).
A rule of thumb
Let tools hold things for you, sort things for you and remind you of things. Be more careful about letting them decide, explain or argue for you, especially when the point is for you to get better at it. If you're using AI to learn, ask it for questions and hints, not answers.
It's worth looking back over a week and sorting what you used AI for into those two piles. Most people find the first pile is bigger than they'd have guessed, and that it's the second pile they'd quite like back.
How TalkGraph splits the work
TalkGraph was designed around exactly this line. It takes the storage and the sorting: it keeps the transcript, turns what you say into a structured map, and remembers every session in your Brain. The thinking stays with you. When you pause, it asks you a question rather than giving you an answer, and when you ask your Brain something, the answer comes from your own words and points to where you said them.
When you do want a draft, a pitch outline or a plan, TalkGraph's agents write it only from what you've said, and mark the gaps rather than filling them with guesses.
TalkGraph maps what you say as you speak and asks you the next question. Free to try, no sign-up.
Klevis is the founder of TalkGraph, which he built and shipped solo. He has a First Class MSc in Computing and Information Systems, and writes about thinking, learning and tools that help people think rather than think for them.
LinkedIn →Sources
- Risko, E. F. & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676-688. doi.org/10.1016/j.tics.2016.07.002
- Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(1), 6. doi.org/10.3390/soc15010006
- Bastani, H. et al. (2025). Generative AI without guardrails can harm learning. PNAS, 122(26). doi.org/10.1073/pnas.2422633122