ChatGPT power words that change how it thinks

How to use these

These are not hidden commands. There is no secret menu. They are shorthand for frameworks and behaviours the model already understands, which means one word can replace a paragraph of explaining what you want.

Add the word to your prompt, or drop it in as a follow up. Nothing here is ChatGPT specific. The same words work in Claude and Gemini, because they are not commands, they are names for ways of thinking.

The one habit that makes these work: use them as a follow up, not an opening. Let it give you the ordinary answer first, then hit it with RED TEAM or BLINDSPOTS. You get the contrast, and the contrast is where the useful part is.

Where they fall down: a word is not a substitute for context. Ask it to RED TEAM a plan it knows nothing about and you get generic objections that could apply to anyone. Give it the real numbers, the real constraint and the real deadline first, then use the word. The framework is only as good as what you feed it.

Learn faster

The group most prompt lists skip, and the one that changes how quickly you pick up something new.

• ELI10 Explains it in plain language, as if to a ten year old. Use ELI5 if you want it simpler again.
• FEYNMAN Makes you explain it simply, which exposes the parts you do not actually understand.
• SOCRATIC Teaches by asking you questions instead of handing you the answer.
• ANALOGY Explains the hard thing using something you already know.
• TEACH BACK You explain it back, and it finds what you got wrong.
• QUIZ ME Turns the topic into recall questions instead of more reading.
• MISCONCEPTIONS Shows what most people get wrong about the subject.
• MENTAL MODEL Gives you a reusable way to think about it, not just an answer to this question.

Try it: Explain this with ELI10, then QUIZ ME on it.

Marketing and business

• ICP Defines exactly who you are for, and who you are not. Short for ideal customer profile.
• JTBD Focuses on the job the customer is hiring you to do, not your feature list. Short for jobs to be done.
• VOC Uses the words your customers actually use, not your internal language. Short for voice of customer.
• POSITIONING Explains why someone picks you over the alternative, including the alternative of doing nothing.

Try it: Write the ICP for this offer, using VOC from the reviews I have pasted below.

Ten more for business

• OBJECTIONS Surfaces the reasons people hesitate, and turns each one into a response you can use.
• MOAT Identifies what a competitor would genuinely struggle to copy.
• GTM Turns an offer into a go to market plan.
• COMPETITOR Reviews your strategy from a rival’s point of view.
• SCALE Names what breaks if demand goes up ten times.
• VALUE PROP Pins down the specific promise you are making.
• PAIN POINTS Surfaces the frustrations your customer would list first, unprompted.
• SEGMENT Splits a broad market into groups worth treating differently.
• MESSAGE-MARKET FIT Checks whether your messaging matches what the audience actually cares about.
• PROOF Names the evidence a sceptical buyer would need before believing you.

Challenge your ideas

This is the section that earns its keep. The model defaults to agreeing with you, and these are how you stop it.

• RED TEAM Attacks your idea the way someone who wants it to fail would.
• PREMORTEM Assumes it has already failed, then works backwards to explain why.
• STEELMAN Builds the strongest possible version of the argument against you.
• FALSIFY Asks what evidence would prove you wrong, and whether that evidence exists.
• BLINDSPOTS Names what you have not considered.
• ASSUMPTIONS Separates what you know from what you are treating as known.
• SKEPTIC Doubts the claims. Asks what the actual evidence is for each one.
• CRITIC Judges the execution. Reviews the work the way an experienced practitioner would.
• DEVIL’S ADVOCATE Argues the other side even when it agrees with you, which is the entire point.
• KILL CRITERIA Defines in advance what would make you stop.

Try it: RED TEAM this launch plan, then run a PREMORTEM.

Think better

• FIRST PRINCIPLES Strips the problem back to what is actually true, instead of what everyone assumes.
• SECOND ORDER Looks past the immediate result to what happens after that.
• INVERSION Asks how you would guarantee failure, then works backwards from there.
• BASE RATES Anchors the answer in what usually happens, not the exceptional story you have in mind.
• MECE Splits a messy problem into categories that do not overlap and do not leave gaps. Short for mutually exclusive, collectively exhaustive.
• FIVE WHYS Asks why repeatedly to get past the symptom and reach the actual cause.
• COUNTERFACTUAL Changes one assumption and shows how differently it plays out.
• SYSTEMS THINKING Looks at how the parts affect each other, rather than analysing them one at a time.
• BOTTLENECK Finds the one constraint holding the whole thing back.
• LEVERAGE Finds the smallest action that moves the most.

Try it: Use FIRST PRINCIPLES to rethink how I price this. Then find the BOTTLENECK.

Make better decisions

• 80/20 Finds the few actions responsible for most of the result.
• OPPORTUNITY COST Shows what you are giving up by choosing this.
• REVERSIBLE Sorts the decisions you can undo cheaply from the ones you cannot.
• DECISION MATRIX Scores every option against the criteria you actually care about.
• TRUE COST Adds time, complexity and risk to the money.
• TRADE-OFFS Makes the downside of each option explicit, instead of selling you all of them.
• EXPECTED VALUE Weighs each option by how likely it is and how much it pays off.
• REGRET MINIMISATION Asks which choice you are least likely to regret later.
• SANITY CHECK A fast pass for obvious flaws and numbers that do not add up.
• NEXT STEP Ends the analysis and gives you one thing to do.

Try it: Build a DECISION MATRIX for these three options, then give me the NEXT STEP.

Better writing

• TRIM Cuts the length without losing the point.
• SPOKEN Rewrites it so it sounds like a person saying it out loud.
• PLAIN Strips the jargon and the long words.
• PUNCH Shorter, sharper, harder to skim past.
• NO FLUFF Kills the throat clearing intro and the generic conclusion.
• VOICE Matches the tone of writing samples you paste in.
• HOOKS Gives you opening lines only, not a rewrite of the whole thing.
• CONTRAST Makes the point land by showing before and after, or old and new.

Try it: TRIM this and make it SPOKEN. NO FLUFF.

The best trick is stacking them

One word is useful. Four in sequence is a different thing entirely, because each one operates on the output of the last.

INSTEAD OF THIS
Is this a good business idea?

TRY THIS
Use FIRST PRINCIPLES to analyse this idea. Then RED TEAM it,
run a PREMORTEM, identify my BLINDSPOTS, and finish with the
80/20 NEXT STEP.

Same idea, same model, same five seconds of typing. The second one comes back with something you can act on, and usually something you did not want to hear.

Write the ICP for this offer and the JTBD it serves. Use VOC
from the reviews below. Then give me the POSITIONING against
the alternative of doing nothing, and the top OBJECTIONS.

Explain this with ELI10. Then give me the MENTAL MODEL, the
common MISCONCEPTIONS, and QUIZ ME on it.

RED TEAM this launch plan. Run a PREMORTEM. List my
ASSUMPTIONS and BLINDSPOTS, set KILL CRITERIA, then give me
the 80/20 NEXT STEP.

Cheat sheet

• Your idea feels obvious. FIRST PRINCIPLES
• You want it to disagree. RED TEAM
• You are about to launch. PREMORTEM
• You are stuck between options. DECISION MATRIX
• You want the fastest path. 80/20
• Something feels off. BLINDSPOTS
• You want it simpler. ELI10
• You are learning it properly. FEYNMAN
• Your writing sounds like AI. SPOKEN, then TRIM, then PLAIN
• You are doing marketing. ICP, JTBD, VOC
• You are overthinking it. NEXT STEP

Common questions

• Are ChatGPT power words real commands? No. There is no secret menu. They are shorthand for frameworks and behaviours the model already understands, which is why one word can replace a paragraph of explaining what you want.
• Do these work in Claude and Gemini? Yes. Nothing here is ChatGPT specific, because these are names for ways of thinking rather than product features. The one exception is Claude Code slash commands, which are covered in the Claude version of this guide.
• What does ELI10 mean? Explain it like I am ten. It asks for plain language with no jargon. Use ELI5 if you want it simpler again, and pair it with QUIZ ME or FEYNMAN when you are trying to actually learn the topic rather than just skim it.
• Why does ChatGPT always agree with me? It defaults to agreeing. RED TEAM, PREMORTEM, STEELMAN, BLINDSPOTS and DEVIL’S ADVOCATE are how you stop it. Use them as a follow up rather than an opening, so you get the contrast between the ordinary answer and the challenged one.
• Should I use a power word at the start or the end of a prompt? As a follow up. Let it give you the ordinary answer first, then apply the word. The contrast between the two is where the useful part is. Opening with the word alone tends to produce generic output.
• Why did my power word give a generic answer? A word is not a substitute for context. Ask it to RED TEAM a plan it knows nothing about and you get objections that could apply to anyone. Give it the real numbers, the real constraint and the real deadline first, then use the word.