Guide

AI prompts for thinking more clearly before you decide

Published 16 October 2026

A chessboard mid-game on a wooden table beside an open notebook with a hand-drawn branching diagram and a pen

Short answerAsk the AI to argue against your own position before you commit to it, not just to agree with you. A model that only confirms what you already think is the least useful one you can build.

Chat AI has a habit that works against you here: it tends to agree with whatever framing you give it. Ask it to back up a decision you've already made, and it usually will. That makes it a weak tool for thinking clearly, unless you deliberately point it the other way and ask it to argue against you, map both sides fairly, or check your reasoning for the patterns that tend to fool people.

None of these prompts make the decision for you. They're built to slow you down for ten minutes before a decision that matters, which is usually enough time to notice the thing you were about to miss.

1. Argue against your own position

Before you commit to a view, ask the AI to make the best possible case for the opposite one. The steelman prompt is built specifically to resist the pull toward agreeing with you:

prompt — 1 blank

I disagree with this position: [POSITION].

Write the strongest, fairest version of it as its most thoughtful supporter would present it. Include the best evidence, the values behind it, and what they think opponents get wrong. Then list the two points where I might be wrong and the two where the position is weakest.

Read the steelman before you argue back, not after. If you can't find a flaw in the strongest version of the other side, that's worth knowing before you act, not after.

Read the steelman before you argue; you will disagree better even if you still disagree.

2. Separate facts from habits and assumptions

A lot of decisions are shaped by "we've always done it this way" rather than by anything that's actually still true. The first principles prompt pulls those two things apart:

prompt — 1 blank

Help me think about [PROBLEM] from first principles.

Ask me what I currently assume, then separate hard facts from assumptions and habits. Rebuild the problem from the facts only, and suggest three solutions I would not have considered if I had followed convention. Question anything that starts with "we have always".

The useful part is being asked what you currently assume, rather than guessing at your own blind spots. Most people can't list their own assumptions without being prompted for them directly; the question itself does most of the work.

3. Map both sides of a genuine disagreement

When you're trying to understand a real argument rather than win one, such as a workplace debate or a family disagreement over a big decision, the argument map prompt lays out both sides without picking a winner:

prompt — 1 blank

Map the main arguments on the question: "[QUESTION OR CLAIM]".

Show the strongest case for each side, the key assumptions behind each, the evidence that would change each side's mind, and where the disagreement is really about values rather than facts. Use neutral language and do not tell me what to think.

The last part, where the disagreement is about values rather than facts, is often the real finding. Two people arguing about a policy or a family decision are frequently not disagreeing about what's true at all; they're weighing the same facts against different priorities, and no amount of extra data will settle that on its own.

4. Imagine the plan has already failed

A pre-mortem is a habit borrowed from project planning, and it works just as well for a personal decision: imagine the plan has already failed, then work backwards to find out why. The pre-mortem prompt does this before you've committed anything:

prompt — 4 blanks

Imagine it is [TIME] from now and my plan has completely failed.

The plan: [PLAN]
Who is involved: [PEOPLE]
Key assumptions: [ASSUMPTIONS]

List the ten most likely reasons it failed, ranked by probability and impact. For each, suggest an early warning sign and one action I can take now to reduce the risk.

Do this while changing course is still cheap, before you've spent money or told other people your plan. The early warning signs it suggests are often more useful than the failure reasons themselves, because they give you something to actually watch for later.

5. Check your own reasoning for bias

The hardest reasoning to check is your own, because the gaps feel invisible from the inside. The bias check prompt asks the AI to read your own explanation back to you and flag the patterns:

prompt — 3 blanks

I am about to decide: [DECISION].

My reasoning: [REASONING]
What I want to happen: [PREFERENCE]

Review my reasoning for cognitive biases such as confirmation bias, sunk cost, overconfidence and anchoring. Quote the parts that show each one, then give me five questions to ask myself and the evidence that would prove me wrong.

Write your reasoning honestly, including the part where you say what you want to happen. Leaving that out makes the check easier to pass and far less useful, since the AI can't flag a bias toward an outcome it doesn't know you want.

Common problems

The AI still seems to agree with you. Make sure you're not restating your conclusion as if it were a fact in the prompt. State the decision and your reasoning separately, and ask explicitly for disagreement, not confirmation.

The output feels too abstract to use. Ask for one concrete next action from each biased pattern it flags, rather than just the label. "You're anchoring on the first number you saw" is less useful than "get a second, independent estimate before you decide."

Someone you trust disagrees with the AI's take. Trust the person. The AI cannot see your real incentives, relationships or history the way someone who knows you can, and it will say so if you ask it to acknowledge that limit.

Where to go from here

If the decision in question is specifically about money, AI prompts for big money decisions covers renting vs buying, big purchases and more with the same structured approach. For practising an argument out loud rather than on paper, AI prompts for language practice includes a debate partner prompt. These prompts live in the chat and writing prompts collection, and the prompt builder can adapt any of them to your own situation.

Quick checklist

Ask for the strongest case against your view, not just support for it.
Separate hard facts from habits and assumptions before you decide.
Map both sides neutrally when you're trying to understand a disagreement, not win it.
Run a pre-mortem while changing course is still cheap.
Write your real preference into a bias check; hiding it makes the check weaker.
Treat every output as a second opinion, not the final word.