20 ChatGPT Prompts for Problem Solving to Think Clearer

Most problems that feel hard are badly defined rather than genuinely difficult. That is why the first four prompts below do not look for a solution at all, and why skipping them is the most common reason the rest disappoint.

The twenty prompts are arranged in five stages: defining the problem, generating options, evaluating them, getting unstuck, and then planning and reviewing. Each has a note explaining why it works, because the reasoning transfers to problems this page has not thought of.

They work in ChatGPT and any comparable model. Four of them - the assumption audit, the pre-mortem, the steel-man and the second-order map - also appear in shorter general-purpose form in our best ChatGPT prompts collection. The versions here are longer and specific to working through a problem in sequence; those are for reaching for one on its own.

• It is good at structure and blind spots. Reframing a problem, listing the assumptions you did not notice you were making, and asking the question you have been avoiding. That is where nearly all the value on this page comes from.
• It does not know your organisation. It cannot know the history, the politics, or the thing that was tried three years ago and failed for reasons nobody wrote down. Supply that context yourself or you will get advice that is sound in general and wrong here.
• A generated decision matrix is not data. Prompt 9 produces criteria and scores, and a table of numbers looks far more rigorous than it is. Those scores are the model's impression, not measurement. Use the criteria; supply your own scores where you have real information.
• Simulated experts are voices, not expertise. Prompt 8 is genuinely useful for surfacing angles you had not considered. It is not a substitute for asking somebody who actually does the job, and it is the prompt on this page most likely to be over-trusted.
• Treat every answer as a hypothesis. Your job is to stress-test the output rather than accept it. Use the model to challenge your thinking, not to confirm it, and notice when you are asking a question you already know the answer you want to.

None of these four asks for a solution. Spending twenty minutes here is the highest-return part of the whole process, and it is the part everybody wants to skip.

Prompt 1: Problem Clarification

I am dealing with the following problem: [describe your problem]. Before we think about solutions, help me define it more precisely. What is the actual problem versus the symptoms I am describing? What would need to be true for this problem to be fully solved? What information am I missing that would change how I approach this? And what is the simplest version of this problem?

Why it works: vague problem statements produce vague solutions. This prompt forces precision before any solution generation begins, which dramatically improves the quality of everything that follows.

Prompt 2: Root Cause Analysis

Here is a problem I keep encountering: [describe the problem]. Help me find the root cause rather than the surface symptom. Apply the Five Whys technique - ask me why five times, drilling deeper with each answer. After each of my answers, ask the next why. At the end, tell me what you believe the root cause is and how confident you are in that diagnosis.

Why it works: the Five Whys is one of the most proven root cause analysis techniques. Running it interactively with AI is more rigorous than doing it in your own head, where you tend to stop at the first plausible-sounding cause.

Prompt 3: Problem Reframing

I have been thinking about this problem in the following way: [describe how you currently frame the problem]. Give me five completely different ways to frame this same situation. For each reframing, explain what new solution space it opens up that my current framing closes off. Include at least one reframing that inverts the problem entirely - asking what would make this problem worse rather than better.

Why it works: the way you frame a problem determines what solutions you can see. This is the most powerful prompt on this list for genuinely stuck thinking - a new frame can make a previously invisible solution obvious.

Prompt 4: Assumption Audit

Here is how I am thinking about this problem: [describe your current understanding and approach]. List all the assumptions I appear to be making - both explicit and implicit. For each assumption, tell me: how likely it is to be true, what evidence would confirm or contradict it, and what would change about my approach if the assumption turned out to be wrong.

Why it works: most failed solutions were built on unexamined assumptions. This prompt surfaces them explicitly and attaches a consequence to each one, making it clear which assumptions are load-bearing and which are harmless.

If you only use one prompt from this page, use prompt 3. A reframe costs five minutes and quite often ends the problem rather than solving it.

Four prompts for widening the field before you narrow it. The discipline that makes them work: generate without evaluating, and resist the urge to dismiss anything until you have the full list.

Prompt 5: Divergent Option Generation

My problem is: [describe problem]. I have already considered the following approaches: [list what you have already thought of]. Generate 10 additional solutions I have not mentioned. Include: at least two options that are unconventional or counterintuitive, at least two that address the problem indirectly, and at least one that eliminates the problem rather than solving it. Do not evaluate the options yet - just generate.

Why it works: separating generation from evaluation is a foundational principle of effective brainstorming. Asking for unconventional and indirect options explicitly prevents the AI from defaulting to the same obvious solutions you have already considered.

Prompt 6: Analogous Problem Search

My problem is: [describe problem]. Find three analogous problems from completely different domains - for example from biology, engineering, military history, game design, or urban planning. Describe how each domain solved its version of the problem and explain what principles from each solution could be translated to my situation.

Why it works: the best solutions to hard problems often come from outside the domain where the problem exists. Looking for structural similarities across fields is how many of the most important innovations were made.

Prompt 7: Constraint Reversal

I am trying to solve [problem] but feel constrained by [list the constraints: budget, time, resources, stakeholder requirements, etc.]. For each constraint, give me two alternative framings: (1) treat the constraint as a creative brief - what solutions become possible only because of this constraint, and (2) question whether the constraint is real or assumed. Which constraints are truly fixed and which might be negotiable?

Why it works: most constraints are softer than they appear. Distinguishing between fixed constraints and assumed ones often reveals more solution space than generating new options within the original boundaries.

Prompt 8: Expert Panel Simulation

My problem is: [describe problem]. Simulate a panel of five experts with different backgrounds - for example a systems thinker, a behavioural economist, a pragmatic engineer, a contrarian, and someone from an adjacent industry. What would each expert say about the most important thing I am overlooking? What solution would each one advocate for and why?

Why it works: this is a structured way to access diverse perspectives without needing a room full of people. The contrarian role is especially valuable - it forces the generation of the best argument against whatever you are currently planning. Treat the output as angles to investigate rather than as expert opinion.

Do not evaluate anything during this stage. The moment you start judging options you stop generating them, and the option you would have thought of eleventh is often the one worth having.

Four prompts for narrowing down. Two of them are designed to attack a decision you have already made, which is uncomfortable and exactly the point.

Prompt 9: Decision Criteria Matrix

I am deciding between the following options: [list options]. Help me build a decision matrix. First, suggest the 5 to 7 most important criteria for evaluating these options given the context: [describe your situation, goals, and constraints]. Then rate each option against each criterion on a scale of 1 to 5 and identify which option scores best overall. Flag any cases where the highest-scoring option might not be the right choice despite its score.

Why it works: decision matrices force you to make your evaluation criteria explicit before you score options, which reduces the bias of working backwards from a preferred conclusion. The final flag is important - quantitative tools can miss qualitative factors that matter. Replace the generated scores with your own wherever you have real information.

Prompt 10: Pre-Mortem Analysis

I am planning to [describe your planned solution or decision]. Conduct a pre-mortem. Imagine it is one year from now and this plan has failed completely. What are the five most likely reasons it failed? For each failure mode, tell me: how probable it is, how severe the consequences would be, and what I could do now to prevent or mitigate it.

Why it works: pre-mortems are one of the most evidence-backed decision-improvement techniques. Imagining failure rather than asking what could go wrong removes the psychological inhibition against identifying problems with a plan you have already committed to.

Prompt 11: Steel-Man the Opposition

I have decided that the best solution to [problem] is [your proposed solution]. Make the strongest possible case against this decision. Do not give me weak objections I can easily dismiss - give me the best arguments a thoughtful, well-informed critic would make. After presenting the steel-man case against my solution, tell me whether any of those arguments should change my decision.

Why it works: confirmation bias is the most dangerous cognitive trap in decision-making. Explicitly requesting the strongest counterargument forces engagement with evidence and reasoning that might genuinely change your mind.

Prompt 12: Second and Third Order Consequences

I am considering this solution: [describe solution]. Map out the second and third order consequences. What are the immediate first-order effects? What do those effects then cause - the second-order consequences? And what do those then produce - the third-order effects? Include both intended and unintended consequences, and flag any cases where a later-order consequence might undermine the first-order benefit.

Why it works: most bad decisions look fine at the first-order level. The problems emerge in second and third-order effects that were not anticipated. Thinking in systems rather than direct cause-and-effect is a skill that separates excellent problem solvers from average ones.

Run prompt 11 after you have decided rather than before. A steel-man argument against a position you have not committed to is an intellectual exercise; against one you have, it is a genuine test.

Four prompts for when you have been circling the same problem for weeks. Each one attacks the situation from a direction that proximity to the problem has closed off.

Prompt 13: The Outsider Perspective

I have been stuck on this problem for [time period]: [describe problem]. Pretend you are someone who has just encountered this problem for the first time with no prior knowledge of my industry, context, or constraints. What are the first three questions you would ask? What would seem obviously strange or inefficient about the situation as I have described it? What obvious solutions might an outsider suggest that an insider would dismiss too quickly?

Why it works: proximity to a problem creates blind spots. The questions a genuine outsider asks, especially the naive ones, often expose assumptions that have calcified into facts for people inside the problem.

Prompt 14: Inversion Thinking

My goal is [describe what you are trying to achieve]. Instead of asking how to achieve this, let's invert the problem. What would I need to do to guarantee I fail? What actions would make this outcome impossible? List the top ten ways to ensure the worst result. Then, for each item on that list, tell me whether I am currently doing any version of it - even unintentionally.

Why it works: inversion, thinking about what to avoid rather than what to do, is one of the most powerful thinking tools available. It surfaces risks and failure modes that forward-thinking misses entirely, and the final check for unintentional failure behaviours is often humbling and actionable.

Prompt 15: Minimum Viable Solution

I am trying to solve [problem] but feel overwhelmed by the complexity of a full solution. What is the smallest, simplest version of a solution I could test in the next 48 hours that would tell me whether I am on the right track? What is the one question this minimal test would answer? And what would I need to see from that test to justify investing in a fuller solution?

Why it works: many problems stay unsolved because the path to a full solution feels overwhelming. Identifying the smallest test that would generate useful signal breaks the paralysis and provides real information rather than continued speculation.

Prompt 16: The Brutal Simplification

Here is my current approach to [problem]: [describe your current thinking in detail]. Simplify this ruthlessly. Strip out everything that is not essential to the core problem. What is the single most important thing I need to decide or do? What could I eliminate, defer, or delegate without significantly changing the outcome? Give me the version of this problem that is 80% as good but 20% as complex.

Why it works: complexity is often a symptom of unclear thinking rather than a genuine feature of the problem. Forcing a simplification exposes what actually matters and frequently reveals that much of the apparent complexity was not load-bearing.

Prompt 14 is the one that stings. Being told which of the ten ways to guarantee failure you are already doing is the most useful unpleasant answer on this page.

The stage where good solutions die. Four prompts for getting a decision implemented, accepted, and then learned from.

Prompt 17: Implementation Risk Assessment

I have decided to implement [solution] to address [problem]. The key steps in my plan are: [outline your implementation plan]. Identify the three points in this plan most likely to go wrong, the warning signs I should watch for at each stage, and what I should do if I see those warning signs. Also tell me what I should do at each stage to maximise the chance of success.

Why it works: implementation is where most good solutions die. Identifying the highest-risk points in a plan before execution, and having contingency thinking in place, dramatically improves the chances of success.

Prompt 18: Stakeholder Objection Mapping

I need to get buy-in from [describe stakeholders: e.g. my manager, the board, the engineering team, sceptical colleagues] for this solution: [describe solution]. For each stakeholder group, tell me: what their primary concern or objection is likely to be, what they care most about that I should address directly, and how I should frame the solution to address their specific perspective. Then tell me the order I should approach them in and why.

Why it works: a good solution that fails to get implemented because it could not get buy-in is a failed solution. Thinking about stakeholder concerns before you walk into the room, and tailoring your framing to each audience, is the difference between persuasion and information delivery.

Prompt 19: After-Action Review

I recently tried to solve [problem] using [approach]. The outcome was: [describe what happened]. Conduct an after-action review. What went as planned and why? What did not go as planned and why? What should I have known or done differently at each stage? What is the single most important lesson to carry into the next similar situation? And what does this reveal about my default problem-solving patterns?

Why it works: the ability to extract generalisable lessons from specific experiences is what separates people who get better at problem-solving over time from those who repeat the same mistakes. The last question, about default patterns, is the most valuable and most overlooked.

Prompt 20: Build Your Problem-Solving Framework

Based on the following description of my role and the types of problems I regularly face: [describe your work and recurring challenges]. Design a personal problem-solving framework for me. It should include: (1) a set of questions to ask at the start of any new problem, (2) a checklist of cognitive biases to watch for, (3) a default process for moving from problem to solution, and (4) the three thinking tools I should have in my toolkit given the specific types of problems I face.

Why it works: having a personal problem-solving process, one you return to consistently, produces better results than approaching each new problem from scratch. This prompt uses the AI to design a framework tailored to your specific context rather than a generic methodology.

Do prompt 19 even when things went well. An after-action review on a success tells you which part was judgement and which part was luck, and that is the only way to know whether to repeat it.

Give the model as much context as you would give a trusted colleague. The more it understands your situation, goals, constraints and what you have already tried, the more useful the output. Everything on this page degrades sharply when the problem is described in three lines.

When you are genuinely stuck, run the same problem through two different models and compare. Disagreements between them mark the places where the answer is not obvious, which is useful information in itself. For the diagnostic prompts, where you want the working shown rather than a conclusion, DeepSeek V4 Pro suits the job. For the divergent ones - ten options, five reframings - a fast tier like Grok 4.5 gets you the volume quickly and you do the selecting.

For prompting technique across other kinds of work, see the main ChatGPT prompts guide. Where a problem turns out to be a data question, our ChatGPT prompts for data analysis go deeper, and the same five-stage structure applied to studying is in our ChatGPT prompts for learning maths. An AI idea generator is useful alongside prompt 5 when you want more volume. Chat Smith is free to try.

And the habit that matters most: use the model to challenge your thinking rather than to validate it. If you find yourself rephrasing a question because you did not like the answer, that is the moment the answer was worth having.