Ten before you prompt: how to avoid over-reliance on AI
Before you ask an AI anything that matters, write ten lines from your own head. Paper or a plain note, no tab open. Then ask. Then put your list next to the answer and compare.
That is the whole rule. It takes about five minutes. You stay the first author of your own thinking, and the AI becomes the second opinion.
What is over-reliance on AI?
Over-reliance on AI is when the tool writes the first draft of your thinking, not only of your text. You stop forming a view before you ask. You read the answer. It sounds right. It becomes your view.
It is not about how much you use AI. You can use it all day and still think for yourself. The question is order. Who goes first: you, or the box?
Why ten lines, and why first?
The number comes from a game, not from a lab.
In 2020 I started my first company. I kept seeing the same thing in the people around me: they looked at the surface of a problem, and then they decided. So I made a small game for my team. Every day at tea time, one person got a random topic. The first one was "cow". They had to say ten points about it, out loud. One rule: the audience said whatever came to their minds first, and the person with the topic could not repeat any of it.
Within two minutes, every shallow thought had been spoken. The person holding the topic still had to find ten. That was hard. But over the weeks, something changed in the team. They thought before they talked. I could see it.
An AI chat is the fastest audience in the world. It says all the easy things in five seconds. If you read it first, the easy things become yours, and you never reach the hard ones. Writing ten first is the same game in the other order. You get to the edge of your own head before anything fills it.
Why ten and not three? Everyone has three. In our game, the easy points ran out long before ten. The work started there.
How to do it: the ten-before-you-prompt rule
- Write your question at the top of a page.
- Write ten numbered lines under it. Short bullets, not sentences. What you know, what you guess, what you doubt, what you would do. No search, no AI.
- When you get stuck, keep going. Ask: who is involved? How does it actually work? What would go wrong? What am I assuming?
- Now open the AI and ask the question the way you normally would.
- Read the answer next to your list. Mark three things: what it said that you also had; what it said that you missed; what you had that it did not.
- Write one last line in your own words: what do I think now?
The last line is the one that matters. It is where you decide, not the tool.
You will not do this for every question. Do it for the ones you will act on: a decision, a plan, a piece of writing with your name on it.
What does the research say about AI and critical thinking?
The short answer: the evidence is early. Nobody knows yet whether using AI makes people worse thinkers over time. Two studies are quoted a lot. Here is what each one looked at.
A survey of knowledge workers (Lee et al., Microsoft Research, CHI 2025). The researchers surveyed 319 knowledge workers, who shared 936 examples of using generative AI at work. Higher confidence in the AI was associated with less critical thinking. Higher confidence in yourself was associated with more. This is what people reported about their own work. It shows a link, not a cause.
"Your Brain on ChatGPT" (Kosmyna et al., 2025). 54 people wrote essays in three groups: with an AI chat, with a search engine, or with no tools. Their brain activity was recorded with EEG. The no-tools group showed the strongest connectivity, and the AI group the weakest. The AI group was also the least able to quote from their own essays. It is a preprint on arXiv. It is one kind of task. Only 18 people came back for the last session.
Neither study shows that AI damages your brain. So the honest answer to "is AI making us dumber?" is: we do not know yet. The rule does not need the answer. It costs five minutes, and you lose nothing if the studies turn out wrong.
Try it for one week
- One real question a day. Something you would have asked an AI anyway.
- Ten lines first. Then the AI. Then the three marks and your last line.
- Keep every page.
- On day seven, read only your "what I had that it did not" lines.
That page is your evidence. Not mine, not a study's. If it is empty, you have learned something too.
What I do not know
I do not know if this works for everyone. I have no numbers from other people yet. What I have is the game, and what I saw it do to my team.
Years later I got stuck myself, in stress and a feed, for a year or two. I built a way out for myself first. This rule sits at the centre of it.
Where the app fits
You do not need an app for this. Paper works.
I built DoomsHuman as the gym for this habit. In every topic, you write ten bullets first, and nothing else in that topic opens until you do. The AI only runs when you press its button, after your ten.
Why build it at all? Not for money. To prepare humans for Dooms Day. When corporate giants or a super AI take over most of the world, a small group of people will rebel. That group will be smart, sharp and real. Writing ten before you prompt is one small step in preparing for that.
If you want the story behind the game, it is on the About page. The game itself, as a practice you can do alone, is in how to think for yourself.
FAQ
Is AI making us dumber?
We do not know yet. Early studies, such as a 2025 survey of 319 knowledge workers and a small 2025 EEG preprint, found links between leaning on AI and less effort in thinking. Neither shows lasting harm. A habit that keeps you thinking first costs little either way.
What is over-reliance on AI?
It is letting the AI form your view before you have one. You ask first, read the answer, and adopt it. The fix is about order, not about using AI less: write your own thoughts first, then ask.
How do I keep my critical thinking skills when I use AI?
Write ten lines from your own head before you prompt. Then compare: what did the AI add, what did it miss, what did you have that it did not? End with one line, in your words, on what you now think. Six more exercises like this are in how to improve critical thinking.
What does "think before you prompt" mean?
It means you form a view of your own before you ask an AI. One concrete version: ten lines on paper first, then the prompt, then a comparison. It takes about five minutes.
