As the late great James Brown belted: talking loud and saying nothing. That’s most prompt lists. The same instructions, wearing different hats. Someone writes one, it does okay, it gets copied a thousand times, and now the whole internet runs on variations of “act as a world-class expert and…” You paste one in and get fluent, nothing back.
The ones below are built to do a job, not to sound impressive. Most prompts ask the model to play the role of an expert and hope something useful comes out. These hand it a task — take a position apart, find the assumption nobody’s challenging, strip copy down to what’s load-bearing. A real question, not a role to perform.
What makes them work isn’t clever. Each one spells out what a bad answer looks like, so the model can’t get away with vague filler. And each one asks it to think, not just produce.
Take them. Change them. Rebuild them around whatever you’re stuck on this week. They run anywhere — ChatGPT, Claude, Dante, whatever tab you’ve got open. The model matters far less than the question you bring it.
POSITIONING
Where brands fool themselves first. These five make it harder.
1. Pressure-test a positioning statement
Here's our positioning: [paste]. Don't tell me what's good about it. Tell me what a sharp competitor would say to take it apart in a pitch. Then tell me what an indifferent customer would say to dismiss it in three seconds. Be ruthless.
2. Find the unowned position
In the [category] market, here are the top five competitors and how they position: [list]. What are they all implicitly agreeing on? What's the assumption none of them are challenging? Give me three positions that would only work if you rejected that assumption.
3. The "what we are not" exercise
For [brand], generate a list of 20 things this brand is NOT. Don't list things it doesn't sell. List the values, behaviors, customer types, and aesthetic choices it actively rejects. The goal is to define the brand by what it refuses.
4. Strip the positioning to one sentence
Here's a 200-word positioning paragraph: [paste]. Compress it to one sentence. The sentence must include exactly one specific verb, one specific noun, and zero adjectives. No abstractions. No marketing language. No "leading" or "innovative."
5. Test against the eulogy standard
If [brand] disappeared tomorrow, what specifically would the world be missing? Not "great products" or "good service." What thing would not exist anymore? Be specific enough that the answer couldn't be true of any competitor.
AUDIENCE
The audience in your deck and the one that actually pays you are rarely the same. These find the gap.
6. The "what they actually want" reframe
Our audience is [description]. Marketers usually describe what these people want as [list]. Tell me what they actually want under that. The deeper need they wouldn't articulate in a focus group. Then tell me how that changes how we'd talk to them.
7. Find the smallest viable audience
For [product/brand], who is the smallest group of people for whom this is a 10/10? Not the broad target. The narrow obsessives. Describe them in vivid, specific detail. What they read, who they trust, what they hate.
8. The audience contradiction test
Our stated target is [audience]. But our actual best customers behave like [different audience]. Reconcile that. Which one should we be designing for, and what's the cost of choosing wrong?
9. Map the rejection
For [brand], who should this brand actively repel? Name the customer type that would dilute the brand if we won them. Then write the line that would make them self-select out.
10. The future audience question
Our current customer is [description]. Who is the customer we'll need in three years if the category shifts the way it's likely to shift? Don't extrapolate. Reason from the change first, then back into who that creates.
MESSAGING
Most copy reads like marketing. These catch it doing that.
11. Strip the marketing voice
Here's a piece of brand copy: [paste]. Rewrite it as if you were explaining it to a smart friend at a dinner party. No selling. No hype. No marketing voice. Just the actual idea, said plainly. Then tell me what the original was hiding.
12. The headline test
Generate 20 headlines for [campaign/product]. After each one, write one sentence explaining what it's making the reader feel and what it's making them want to do next. If you can't articulate that clearly, the headline is wrong.
13. Find the unsayable thing
For [brand/category], what is the true thing that nobody is allowed to say out loud in marketing? The thing that would actually move people but feels too risky, too direct, too uncomfortable. Now tell me how we could say it.
14. The competitor mirror
Read the messaging on [competitor]'s homepage: [paste]. Identify the structure they're using. The implicit promise. The proof. The call to action. Now write our message using the OPPOSITE structure. What do we get?
15. Cut the verbs
Here's our copy: [paste]. List every verb in it. Identify the ones that are doing real work and the ones that are filler ("enable," "deliver," "provide," "leverage"). Rewrite using only the working verbs.
NAMING
AI gives you a hundred names and zero judgment. These force the judgment.
16. Name from a feeling
I want to name a [product]. The feeling I want it to evoke is [adjective + adjective]. Don't give me names that describe the product. Give me 30 names that sound like that feeling. Pull from any source. Nature. Mythology. Music. Sports. Whatever.
17. The phonetic gut-check
Here are five name candidates: [list]. For each, tell me what the sound of the name evokes. Hard consonants vs. soft. Length. Rhythm. Then tell me which one would be hardest for a competitor to mock.
18. Names by category violation
For a [category] product, the standard naming convention is [pattern]. Give me 20 names that deliberately violate that convention. Names that would feel like they belong in a different category entirely.
19. The trademark stress test
Here's a name: [name]. Brainstorm every reason this name could fail. Phonetic confusion. Negative connotations in other languages. Trademark conflicts. Awkward URL. Unfortunate abbreviation. Be exhaustive.
20. Stress-test the wordmark
Here's a brand name: [name]. Imagine it set in three different typographic styles. A serious editorial serif. A brutalist sans. A soft humanist serif. In each context, what would the name communicate? Which one contradicts the brand’s intent?
STRATEGY AND DECISIONS
For when a plan feels too comfortable.
21. The "what would we do if we were brave" prompt
Here's our current plan: [paste]. Tell me what this plan would look like if we cut all the safety hedges and trusted our hypothesis fully. What would we do differently? What's the cost of NOT being that brave?
22. Find the load-bearing assumption
Our strategy is [paste]. Identify the single assumption this entire strategy rests on. If that assumption turned out to be wrong, what falls apart? Now: how confident are we in that assumption, and what would prove it?
23. The pre-mortem
It's 12 months from now. [Project] has failed. Tell the story of how it failed. The warning signs we ignored. The decisions that compounded badly. The moment it became obvious. Be specific.
24. The opposite play
Our strategy is [paste]. Now write the strategy memo for the exact opposite approach. Not "different." Opposite. What would have to be true for that approach to be correct? Is any of it true?
25. Force a tradeoff
For [brand/product], we're trying to be: [list values]. That list is too long. Force the tradeoffs. Which two of those, if we doubled down on them, would require us to actively give up the other three? Which trade is most defensible?
SHARPENING
Last set. Run these on anything before it goes out.
26. The "say it like you mean it" rewrite
Here's the draft: [paste]. Rewrite it removing every instance of: hedging language, qualifying phrases, "we believe," "potentially," "could help," and any adverb ending in -ly. Then tell me what's left.
27. Read it as your harshest critic
Read [paste] as the most cynical, intelligent reader in your audience. What's the first eye-roll moment? What's the line that makes them stop reading? Where does the writer lose them?
28. The screenshot test
Pull the single most quotable sentence from [paste]. If someone screenshotted that line out of context and posted it, would it land? If not, what sentence SHOULD be in there to pass the test?
29. The boring-on-purpose test
For [headline/tagline], rewrite it as boringly as possible. Flat. Descriptive. Zero personality. Now tell me what the original was actually adding versus what was performative. Sometimes boring is more honest.
30. The "would a human send this" check
Read [email/post/copy]. If a real person on your team sent this exact text from their personal account, would it sound like them, or would it embarrass them? If it would embarrass them, tell me which specific phrases gave it away as marketing-voice.
HOW TO USE THESE
These work best when you already have something to react to — a positioning statement, a draft, a name you’re unsure about. Bring that, then use the prompts to find the weak spots before someone else does. They won’t build the strategy for you. They test what you already have.
The AI isn’t the strategist. You are.
Copy them, change them, write your own. They get more useful the more you adapt them to real work.