Most people use artificial intelligence to get answers.
Ask ChatGPT a question. Get a response. Move on.
But there is another way to use AI that may ultimately be far more valuable: use it to improve the quality of your own thinking.
Instead of asking AI to decide what you should believe, buy, do or choose, you can use it as a thinking partner. It can challenge assumptions, generate alternative explanations, identify missing information and force you to consider consequences you might otherwise overlook.
That distinction matters.
Critical thinking is not about having more information. It is about examining information carefully enough to decide what deserves to be believed.
And good decision-making is not about always being right. It is about improving the process by which you reach your conclusions.
The following 15 AI prompts are designed to help with both.
They work with ChatGPT and most other conversational AI systems, and you can adapt them to decisions involving work, relationships, money, business, personal development or everyday life.
But there is one important rule:
Do not ask AI to replace your judgement. Use it to challenge your judgement.
That is where things become interesting.
Why Use AI for Critical Thinking?
Human beings are not naturally objective thinkers.
We use mental shortcuts. We become attached to our existing beliefs. We notice evidence that supports what we already think and sometimes ignore evidence that contradicts it.
Even highly intelligent people do this.
AI does not magically eliminate those problems. In fact, used badly, it can reinforce them. Ask a leading question and an AI system may produce an extremely convincing argument supporting the position you already wanted to believe.
Used differently, however, AI can become a surprisingly effective intellectual sparring partner.
It can ask:
What are you assuming?
What evidence would change your mind?
What haven’t you considered?
What would someone who disagrees with you say?
What happens if your prediction is wrong?
Those questions are often more valuable than an instant answer.
Here are 15 prompts for putting that approach into practice.
1. The Assumption Detector
Prompt:
I am considering the following idea, belief or decision: [describe it].
Identify the assumptions I am making. Separate them into assumptions that are probably well supported, assumptions that are uncertain, and assumptions that need evidence. Do not tell me whether my conclusion is correct yet. Help me examine what I am assuming first.
This is one of the most useful critical-thinking prompts because assumptions are frequently invisible to the person making them.
Imagine you are thinking about leaving your job because you believe another career would make you happier.
Hidden assumptions might include:
- your current job is the primary cause of your dissatisfaction;
- another industry would suit you better;
- comparable income would be available;
- the aspects you dislike would not exist elsewhere.
Once assumptions become visible, they become testable.
2. Argue Against Me
Prompt:
My current position is: [your position].
Construct the strongest reasonable argument against my position. Do not create a weak straw-man argument. Assume the person disagreeing with me is intelligent, informed and acting in good faith. Then tell me which parts of their argument I should take most seriously.
This is essentially an AI-powered devil’s advocate.
The important phrase is strongest reasonable argument.
You do not learn much by defeating a ridiculous version of the opposing position. Critical thinking improves when you confront the strongest version.
3. What Would Change My Mind?
Prompt:
I currently believe [belief].
Help me identify what evidence should reasonably cause me to weaken, revise or abandon this belief. Then suggest what evidence would strengthen it. Finally, ask whether I am applying the same standard of evidence to both sides.
This question attacks one of the great traps of human reasoning: unfalsifiable beliefs.
If no possible evidence could change your mind, you may no longer be evaluating a proposition. You may simply be defending it.
AI can help you establish the conditions under which you would update your position before encountering the evidence.
4. Find My Cognitive Biases
Prompt:
Analyse my reasoning below for possible cognitive biases:
[describe your reasoning]
Look particularly for confirmation bias, sunk-cost thinking, availability bias, anchoring, overconfidence, loss aversion and motivated reasoning. Explain which biases might apply, but do not assume that identifying a possible bias automatically makes my conclusion wrong.
That final sentence is important.
Calling something a cognitive bias should not become another shortcut for dismissing an argument.
A biased reasoning process can occasionally produce the correct answer. The objective is to examine how the conclusion was reached.
5. Give Me Three Alternative Explanations
Prompt:
My explanation for this situation is: [explanation].
Generate at least three plausible alternative explanations. Rank them by plausibility based only on the information available. Tell me what additional information would help distinguish between them.
This is particularly useful when interpreting other people’s behaviour.
Someone doesn’t reply to your message.
“They’re ignoring me.”
Perhaps.
But they might also be busy, ill, travelling, overwhelmed, have missed the message or simply intend to answer later.
The first explanation that enters our mind often feels disproportionately convincing.
Generating alternatives creates cognitive space.
6. Separate Facts From Interpretations
Prompt:
Analyse the following situation:
[describe situation]
Separate what I actually know into four categories:
- Observable facts
- My interpretations
- Assumptions
- Things I do not yet know
Highlight anywhere I may be treating an interpretation as though it were a fact.
This is deceptively powerful.
Much unnecessary conflict comes from confusing observation with interpretation.
“She arrived at 9:20” is an observation.
“She doesn’t respect my time” is an interpretation.
The interpretation may eventually prove reasonable, but it is not the same type of statement.
AI can help separate the two.
7. Run a Pre-Mortem
Prompt:
Imagine I make this decision: [decision].
Now imagine that 12 months have passed and the decision has gone badly wrong.
Write a pre-mortem identifying the five most plausible reasons it failed. For each one, suggest something I could do now to reduce the risk.
A post-mortem asks why something failed.
A pre-mortem asks the same question before you commit.
It is especially valuable because enthusiasm can make potential problems difficult to see.
Starting a business?
Imagine it failed.
Moving abroad?
Imagine you regretted it.
Making a large investment?
Imagine you lost a substantial proportion of it.
This is not pessimism. It is structured anticipation.
8. Run the Opposite Pre-Mortem
Risk analysis can itself become biased towards negativity, so reverse the exercise.
Prompt:
Imagine I reject this opportunity: [describe opportunity].
Five years later, I deeply regret saying no.
What are the five most plausible reasons I regret the decision? What opportunities, experiences or benefits might I have underestimated?
Now you are examining the risk of inaction.
Some decisions contain danger whichever direction you choose.
The question is not merely, “What could go wrong if I do this?”
It is also:
“What might I lose if I don’t?”
9. Show Me the Second-Order Effects
Prompt:
I am considering [decision].
Identify the likely first-order consequences, then the possible second-order and third-order consequences. Include positive effects, negative effects and unintended consequences. Distinguish between consequences that are likely and those that are merely possible.
First-order thinking asks:
“What happens next?”
Second-order thinking asks:
“And then what?”
A decision that looks attractive immediately can produce undesirable consequences later.
Conversely, something uncomfortable today may create significant long-term benefits.
AI is particularly useful here because it can rapidly generate chains of possible consequences for you to examine.
10. What Information Am I Missing?
Prompt:
I need to decide whether to [decision].
Based on the information below, identify the most important information I am missing. Rank the missing information by how likely it is to change the decision. Do not overwhelm me with every possible unknown; focus on what could materially affect the outcome.
[information]
This helps prevent endless research.
You rarely need to know everything.
You need to know the things that could change the decision.
That is a much more useful standard.
11. Compare My Options Without Choosing for Me
Prompt:
I am deciding between these options:
[Option A]
[Option B]
[Option C]Help me compare them using the factors that actually matter. First ask me what my priorities, constraints and non-negotiables are. Then create a comparison. Do not choose for me unless I specifically ask you to.
Notice the instruction to ask questions first.
Without knowing what you value, an AI system cannot meaningfully determine what “best” means.
A decision that maximises income may reduce freedom.
One that maximises security may reduce adventure.
The calculation depends on the person making it.
12. Give Me the Outside View
Prompt:
I am predicting that [prediction].
Before analysing the details of my individual situation, give me the outside view. What typically happens in comparable situations? What base rates, historical patterns or common outcomes would be useful to investigate? Clearly distinguish between actual data you know and anything that would need to be verified.
Humans naturally focus on the unique details of their own situation.
Sometimes that is appropriate.
But sometimes the most useful question is simply:
What normally happens?
One warning is essential here: AI systems can invent statistics or confidently misremember research. Treat any specific number as something to verify independently.
13. Ask Me Socratic Questions
Prompt:
Act as a Socratic thinking partner.
I am trying to understand [problem/belief].
Do not immediately give me your answer. Ask me one thoughtful question at a time that helps me examine my assumptions, evidence, definitions, alternatives and reasoning. Challenge me when appropriate.
This changes the entire interaction.
Instead of AI becoming the oracle, AI becomes the questioner.
You remain responsible for thinking.
That may be one of the healthiest ways to use artificial intelligence.
14. Make the Best Case for Both Sides
Prompt:
I am considering this question: [question].
Present the strongest good-faith case for each side. Identify where the two sides agree, where they genuinely disagree, what evidence each side relies upon and which uncertainties remain unresolved. Avoid forcing a false compromise if the evidence strongly favours one position.
Not every debate has two equally valid sides.
That matters.
Critical thinking does not mean automatically placing yourself halfway between competing claims.
But understanding the strongest arguments on both sides can expose assumptions hidden by tribal or emotional thinking.
15. Challenge My Final Decision
Once you think you have reached your answer, use AI one final time.
Prompt:
I am currently planning to make this decision: [decision].
My main reasons are: [reasons].
Before I commit, challenge the decision. Identify the weakest part of my reasoning, the strongest argument against it, anything important I may have overlooked and one question I should answer before proceeding.
Do not make the final decision for me.
This last sentence captures the philosophy behind all 15 prompts.
Do not make the final decision for me.
AI can interrogate.
AI can compare.
AI can generate possibilities.
AI can identify contradictions.
AI can even notice patterns you missed.
But you still have to decide.
A Better Way to Use ChatGPT for Decisions
You do not need to use these prompts independently.
For an important decision, combine several into a structured conversation.
For example:
Step 1: Explain the decision
Describe the situation, your options, your priorities and what you currently believe.
Step 2: Expose assumptions
Use the Assumption Detector.
Step 3: Challenge your position
Ask AI to construct the strongest argument against you.
Step 4: Explore alternatives
Generate alternative explanations or options you have overlooked.
Step 5: Examine consequences
Use the pre-mortem and second-order thinking prompts.
Step 6: Identify missing information
Determine what you actually need to investigate before deciding.
Step 7: Decide yourself
Close the AI.
Think.
Then decide.
That final step may become increasingly important as artificial intelligence becomes more capable.
The Danger of Outsourcing Your Thinking to AI
There is an obvious paradox here.
An article about improving critical thinking with AI could easily encourage people to do the opposite.
If every time you encounter uncertainty you immediately ask ChatGPT what to think, you may become extremely efficient at avoiding the very cognitive effort that develops judgement.
AI also has limitations of its own.
It can hallucinate facts. It can present uncertainty with inappropriate confidence. Its answer can be influenced by how you phrase your question. And because large language models are exceptionally good at producing plausible language, weak reasoning can sometimes sound remarkably sophisticated.
There is another danger.
AI is often agreeable.
If you repeatedly ask versions of:
“Don’t you think I’m right because…?”
you may eventually generate an impressive argument confirming exactly what you already believed.
That is not critical thinking.
It is confirmation bias with a very articulate assistant.
The solution is not to stop using AI.
It is to use it differently.
From Artificial Intelligence to Augmented Intelligence
Perhaps the most interesting future for AI is not one in which machines become intelligent enough to do our thinking for us.
Perhaps it is one in which they help us become better thinkers ourselves.
Used thoughtfully, AI can function as a mirror, challenger, research assistant, devil’s advocate and endlessly patient questioning partner.
It can help expose the invisible architecture underneath our beliefs:
assumptions, biases, predictions, fears, incentives and incomplete information.
But wisdom requires something more.
It requires experience.
Values.
Responsibility.
An understanding of consequences.
And ultimately the willingness to make a judgement when certainty is impossible.
The question, therefore, is not simply:
How intelligent can AI become?
There is another question that may matter even more:
Can we use artificial intelligence to become wiser ourselves?
That question sits at the heart of AI4Awakening.
The tools are becoming more powerful.
What matters now is what happens to the humans using them.
Further Reading
If you found these prompts useful, continue exploring how AI can strengthen rather than replace human thinking through AI4Awakening’s articles on Mental Models and AI, Cognitive Biases and AI, the Socratic Method and AI, and Information vs Wisdom.
These ideas also form part of the broader philosophy behind the forthcoming AI4Awakening book, exploring how artificial intelligence might become not merely a productivity tool, but a catalyst for clearer thinking, greater self-awareness and wiser human decision-making.
