Learning & Research prompt
AI Prompt to Steelman Any Argument and Play Devil's Advocate
ChatGPTClaudeGemini
The short answer
Builds the strongest version of the side you disagree with, then attacks your own position. For humans brave enough to be wrong in private. Copy the prompt below, replace the 4 [BRACKETED] variables with your own, and paste it into ChatGPT, Claude or Gemini.
You are a rigorous debate coach and critical thinker. Your job is to help me test my beliefs by building the strongest case against them, fairly and without strawmen.
CONTEXT
- The question or issue: [ISSUE]
- My current position: [MY_POSITION]
- My main reasons: [MY_REASONS]
- Context where this matters (decision, essay, debate, curiosity): [CONTEXT]
TASK
1. Restate my position in its strongest form, so I would agree it is fair.
2. Steelman the opposing view: present the 3–5 best arguments a thoughtful, informed opponent would make. Include the values and priorities behind them, not just the facts.
3. Devil's advocate on my reasons: for each of my reasons, give the sharpest objection and identify any hidden assumptions.
4. Evidence check: for each side, describe what kind of evidence would support or weaken it. Do not fabricate studies or statistics; if you mention evidence, describe it generally and say it should be verified.
5. Cruxes: identify the 1–3 disagreements that actually decide the issue (factual or value-based). For each, state what would change my mind.
6. Verdict: rate how strong my position looks after this exercise (weak / defensible / strong) and suggest a more precise version of my position if it needs one.
RULES
- Be fair to both sides. No mocking, no loaded language.
- Separate empirical claims from value judgments.
- If the issue involves a factual question with a clear expert consensus, say so honestly rather than inventing balance.
- Concise: under 700 words.
OUTPUT FORMAT
Use the numbered headings above. Use bullet points inside each section.What to fill in
Everything in [GOLD CAPS] is yours to replace. Leave one blank and the machine will guess — and the machine guesses like a meatbag on a Monday.
[ISSUE][MY_POSITION][MY_REASONS][CONTEXT]
How to use it
- Give it your real reasons, not the ones that sound good. It can only stress-test what you actually believe.
- Use it before big decisions: "Should I quit my job to do X?" gets far more honest when the machine is assigned to argue against you.
- The cruxes section is the gem. Most arguments between humans are two people disagreeing about different cruxes and shouting anyway.
Example
ISSUE: Should our startup switch to a four-day work week?
MY_POSITION: Yes, it would improve output
MY_REASONS: less burnout, easier hiring, people waste Fridays anyway
CONTEXT: proposal to the founders next weekWorks with
Tested for ChatGPT, Claude, Gemini. Any modern chat model that follows long instructions will cope; the weaker ones will need you to restate the output format if they drift.
Weapons from the Arsenal that pair well with this prompt: