25 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 five prompts below do not look for a solution at all, and why skipping them is the most common reason the rest disappoint.

The 25 ChatGPT prompts for problem solving below are arranged in five stages: defining the problem, generating options, evaluating them, getting unstuck, and planning and reviewing. They work in ChatGPT and any comparable AI chatbot.

Four of them — the assumption audit, the pre-mortem, the steel-man and the second-order map — also appear in shorter form in our best ChatGPT prompts collection. The versions here are built to be used in sequence.

• Use it for structure and blind spots. Reframing a problem and surfacing hidden assumptions is where most of the value lies.
• It does not know your organisation. Supply the history, politics and past attempts yourself.
• A generated score is not data. Keep the criteria from prompt 11, but replace the scores with your own where you have real information.
• Simulated experts are voices, not expertise. Treat prompt 9 as angles to check with people who actually do the job.
• Treat every answer as a hypothesis. Use the model to challenge your thinking, not to confirm it.

None of these five asks for a solution. Twenty minutes here is the highest-return part of the whole process. DeepSeek V4 Pro suits these diagnostic prompts, where you want the working shown rather than a quick conclusion.

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 it to be fully solved? What information am I missing that would change my approach? And what is the simplest version of this problem?

2. Root cause analysis

Here is a problem I keep encountering: [describe the problem]. Help me find the root cause rather than the symptom using the Five Whys: ask me why, wait for my answer, then ask the next why, five times in total. At the end, tell me what you believe the root cause is and how confident you are.

3. Problem reframing

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

4. Assumption audit

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

5. Problem statement writer

Turn everything we have discussed into a one-paragraph problem statement: who is affected, what is happening, the impact, the likely root cause and what a good outcome looks like — without proposing a solution.

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

Five prompts for widening the field before you narrow it. Generate without evaluating. A fast model like Grok 4.5 gets you volume quickly while you do the selecting.

6. Divergent option generation

My problem is: [describe problem]. I have already considered: [list approaches]. Generate 10 additional solutions, including at least two that are unconventional, two that address the problem indirectly and one that eliminates the problem rather than solving it. Do not evaluate them yet — just generate.

7. Analogous problem search

My problem is: [describe problem]. Find three analogous problems from completely different fields — such as biology, engineering, game design or urban planning. Describe how each field solved its version and what principles I could translate to my situation.

8. Constraint reversal

I am trying to solve [problem] but feel constrained by [budget, time, resources, stakeholder requirements]. For each constraint, show me what becomes possible only because of it, and question whether it is real or assumed. Which constraints are truly fixed and which might be negotiable?

9. Expert panel simulation

My problem is: [describe problem]. Simulate a panel of five people with different backgrounds — a systems thinker, a behavioural economist, a pragmatic engineer, a contrarian and someone from an adjacent industry. What would each say I am overlooking, and what solution would each advocate?

10. Hybrid solution builder

Here are the most promising options so far: [paste]. Combine elements of them into two hybrid solutions, and explain what each hybrid keeps, what it drops and why it might work better than any single option.

Do not evaluate anything during this stage — the option you would have thought of eleventh is often the one worth having. For even more volume, the AI idea generator works well alongside prompt 6.

Five prompts for narrowing down, two of which are designed to attack a decision you have already made. Claude Sonnet 5 tends to give careful, balanced critiques.

11. Decision criteria matrix

I am deciding between: [list options]. Given my situation, goals and constraints — [describe] — suggest the 5 to 7 most important criteria, rate each option against each on a scale of 1 to 5, and flag any case where the highest-scoring option might still be the wrong choice.

12. Pre-mortem analysis

I am planning to [describe your decision]. Imagine it is one year from now and the plan has failed completely. What are the five most likely reasons? For each, tell me how probable it is, how severe it would be and what I could do now to prevent it.

13. Steel-man the opposition

I have decided that the best solution to [problem] is [your solution]. Make the strongest possible case against it — the arguments a thoughtful, well-informed critic would make, not easy objections. Then tell me whether any of them should change my decision.

14. Second- and third-order consequences

I am considering this solution: [describe]. Map its first-, second- and third-order consequences, intended and unintended, and flag any case where a later effect might undermine the first-order benefit.

15. Reversibility check

For each option I am considering — [list] — tell me how reversible it is and the cost of being wrong, and whether I should decide quickly and adjust or slow down and gather more information first.

Run prompt 13 after you have decided rather than before — against a position you have committed to, it is a genuine test. For decisions at work specifically, see our Claude prompts for business.

Five prompts for when you have been circling the same problem for weeks. GPT-5.6 Sol is useful when you want a structured, step-by-step breakdown.

16. The outsider perspective

I have been stuck on this problem for [time period]: [describe problem]. Pretend you have just encountered it with no knowledge of my industry or constraints. What three questions would you ask first, what would seem strange or inefficient, and what obvious solutions might an insider dismiss too quickly?

17. Inversion thinking

My goal is [describe goal]. Instead of asking how to achieve it, list the ten surest ways to guarantee failure. Then, for each, tell me whether I am currently doing any version of it, even unintentionally.

18. Minimum viable solution

I feel overwhelmed by the complexity of solving [problem]. What is the smallest version of a solution I could test in the next 48 hours, what one question would it answer, and what result would justify investing in a fuller solution?

19. Brutal simplification

Here is my current approach to [problem]: [describe in detail]. Simplify it ruthlessly. What is the single most important thing I need to decide or do, what could I eliminate, defer or delegate, and what is the version that is 80% as good but 20% as complex?

20. Explain it back

Ask me one question at a time about [problem] as if you knew nothing about it, until I can explain the problem and my next step clearly in two sentences.

Prompt 17 is the one that stings — being told which failure behaviours you already show is the most useful unpleasant answer on this page. For the same staged approach applied to studying, see our ChatGPT prompts for learning maths.

The stage where good solutions die. Five prompts for getting a decision implemented, accepted and learned from. When the next step is persuading someone in writing, the AI email writer helps you draft the message.

21. Implementation risk assessment

I have decided to implement [solution] to address [problem]. The key steps are: [outline plan]. Identify the three points most likely to go wrong, the warning signs to watch for at each, what to do if I see them, and how to maximise the chance of success at each stage.

22. Stakeholder objection mapping

I need buy-in from [stakeholders] for this solution: [describe]. For each group, tell me their likely objection, what they care about most and how to frame the solution for them. Then suggest the order to approach them in and why.

23. Action plan builder

Turn my chosen solution — [describe] — into an action plan with steps, owners, deadlines, the first action for tomorrow and how we will measure progress.

24. After-action review

I recently tried to solve [problem] using [approach], and the outcome was: [describe]. What went as planned and why, what did not and why, what should I have done differently, what is the single most important lesson, and what does this reveal about my default problem-solving patterns?

25. Personal problem-solving framework

Based on my role and the problems I regularly face — [describe] — design a personal problem-solving framework: questions to ask at the start of any problem, a checklist of cognitive biases to watch for, a default process from problem to solution, and the three thinking tools I should keep in my toolkit.

Do prompt 24 even when things went well — it tells you which part was judgement and which was luck. For turning plans into schedules and status updates, see our ChatGPT prompts for project management.

Give the model as much context as you would give a trusted colleague: your situation, goals, constraints and what you have already tried. 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 models and compare — disagreements mark the places where the answer is not obvious. If a problem turns out to be a data question, our ChatGPT prompts for data analysis go deeper, and for prompting technique in general, see the main ChatGPT prompts guide.

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