THE 18 BEST STARTUP FOUNDER PROMPTS
A model will happily tell you your startup idea is brilliant. These eighteen prompts do the harder, more valuable thing — they try to kill the idea, pressure the assumptions, and force the validation most founders skip.
By Editorial · Published Jun 25, 2026 · 17 min read
On This Page
Startup prompts are usually the most flattering and least useful AI content a founder can find. Ask a model whether your idea is good and it will produce a structured, confident, genuinely encouraging analysis of almost anything — including ideas that should be killed on contact. That encouragement feels like progress and is the opposite of it, because the one thing an early-stage founder most needs is not validation but a fast, honest attempt to find out why the idea will fail. Used well, AI is the cheapest reality-check engine ever built: a tireless sparring partner that will argue against your idea, expose the assumptions you are quietly leaning on, and design the experiments that produce real evidence. This library is eighteen prompts built to do that work, written out in full with no placeholders.
This is a working resource for founders who would rather kill a bad idea this week than next year. Every prompt below is complete and ready to paste; you supply your idea and your context — and the willingness to let the model attack what you have built.
How these prompts are built
Every prompt here follows the same shape, and for founders that shape exists to fight the model's instinct to cheer you on. Each one assigns a role (a skeptical investor, a customer-discovery expert, a red-team analyst), supplies the context of your idea and situation, names the exact deliverables, and imposes constraints — the most important being a bias toward finding what is wrong, and an absolute ban on fabricating traction or market numbers. A prompt that confirms your idea has wasted your time; one that finds the fatal flaw before you spend a year on it has paid for itself a thousand times over.
The prompts run on a small set of variables. Replace these before running any prompt.
| Variable | Replace with | Example |
|---|---|---|
[IDEA] | Your startup or product idea | An AI bookkeeper for trades |
[CUSTOMER] | Who you think you serve | Solo electricians and plumbers |
[PROBLEM] | The pain you claim to solve | Hours lost to manual invoicing |
[STAGE] | Where you are now | Pre-launch, no customers |
[GOAL] | What you are trying to learn or do | Validate willingness to pay |
These tokens are intentional fill-ins, not unfinished sections. The eighteen prompts are grouped into five stages — pressure-test the idea, understand the customer, build the smallest test, make the case, and grow. Worked this way, AI sharpens the judgment behind the things that actually determine startup outcomes: real demand, economic moats, network effects, and the credibility a raise from venture capital demands.
Stage 1 — Pressure-test the idea
Before anything else, try to kill it. These four prompts attack the idea and the problem behind it, because the fastest path to a good startup is eliminating the bad version of it quickly.
1. Steelman and kill
This prompt argues the strongest case for your idea and then attacks it just as hard, so you see both the real opportunity and the real risks. The kill half is the part that matters. If the idea survives an honest assault, you have learned something; if it does not, you just saved a year.
You are an investor who first steelmans an idea, then attacks it without mercy.
CONTEXT
- The idea: [IDEA].
- The stage: [STAGE].
TASK
Make the best case, then the strongest case against.
DELIVERABLES
1. The steelman: the most compelling version of why this could be big.
2. The three most serious reasons it could fail, ranked by how fatal each is.
3. The single assumption that, if wrong, kills the whole thing.
4. A blunt verdict: is this worth pursuing, killing, or reshaping - and why.
CONSTRAINTS
- Attack the real idea, not a strawman; but do not soften the attack to be kind.
- No encouragement for its own sake; I need the truth, not a morale boost.
- If a fatal flaw exists, lead with it.
2. Problem validator
This prompt interrogates whether the problem you are solving is real, painful, and frequent enough to build on. A weak problem dooms even a brilliant solution. It separates a genuine pain from a nice-to-have you have talked yourself into.
You are a customer-discovery expert testing whether a problem is real.
CONTEXT
- The problem I claim to solve: [PROBLEM].
- Who has it: [CUSTOMER].
TASK
Pressure-test whether this is a problem worth solving.
DELIVERABLES
1. How painful, frequent, and expensive this problem plausibly is - and what would confirm it.
2. How people solve it today, including workarounds and doing nothing.
3. Whether this looks like a must-solve problem or a nice-to-have.
4. The questions I should ask real customers to confirm the pain exists.
CONSTRAINTS
- Distinguish a painful problem from a mild inconvenience.
- Do not assume the problem is real because I framed it that way.
- Treat existing workarounds as serious competition.
3. Why-now check
This prompt tests whether there is a real reason this idea works now and not five years ago or five years from now. Timing kills more startups than ideas do. A missing "why now" is often a sign the window is closed or not yet open.
You are a market analyst testing the timing of an idea.
CONTEXT
- The idea: [IDEA].
TASK
Assess whether now is the right time.
DELIVERABLES
1. The "why now" case: what has recently changed that makes this possible or necessary.
2. Whether that shift is real and durable or hype.
3. Why this has not already been done - and whether that reason still holds.
4. The risk that I am too early or too late, and the signal that would tell me.
CONSTRAINTS
- Be honest if there is no compelling "why now"; that is a finding, not a gap to fill.
- Distinguish a real enabling shift from a passing trend.
- Do not invent market events; flag what I should verify.
4. Riskiest-assumption finder
This prompt surfaces the beliefs your whole venture rests on and ranks them by how dangerous they are if wrong. It tells you what to test first. Founders usually pour effort into building before validating the assumption most likely to be false.
You are an analyst finding the assumptions that could sink my startup.
CONTEXT
- The idea and plan: [IDEA].
- The stage: [STAGE].
TASK
Expose what has to be true for this to work.
DELIVERABLES
1. The core assumptions, sorted into facts, reasonable beliefs, and hopes.
2. The assumption that is both most uncertain and most damaging if wrong.
3. The order in which I should test them, riskiest first.
4. The cheapest test for the single riskiest assumption.
CONSTRAINTS
- Be ruthless about which "facts" are really assumptions.
- Rank by uncertainty times impact, not by what is easiest to test.
- Do not let a comfortable assumption pass unexamined.
Stage 2 — Understand the customer
Startups die from building something nobody wants. These four prompts force you toward the real customer and real demand, using AI to sharpen your questions rather than answer them for you.
5. Customer-discovery guide
This prompt designs an interview guide that gets honest signal instead of polite encouragement, avoiding the leading questions that produce false validation. The quality of your discovery depends entirely on not asking people whether they like your idea. It builds questions about their life, not your product.
You are a customer-discovery expert designing an unbiased interview guide.
CONTEXT
- Who I want to interview: [CUSTOMER].
- What I need to learn: [GOAL].
TASK
Design questions that produce honest signal.
DELIVERABLES
1. Open questions about the customer's actual behavior and past experience, not my idea.
2. Questions that uncover the real problem and how they handle it today.
3. The leading questions to avoid, and why they produce false positives.
4. How to listen for genuine pain versus politeness.
CONSTRAINTS
- Focus questions on the customer's life, not on reactions to my solution.
- Avoid anything that invites the interviewee to be nice rather than honest.
- Prioritize learning over selling.
6. Beachhead definer
This prompt narrows a broad market down to the specific first customer you can actually win, which is almost always smaller and sharper than founders want. Trying to serve everyone at the start serves no one. It forces a real beachhead.
You are a go-to-market strategist defining my beachhead customer.
CONTEXT
- The broad market I imagine: [CUSTOMER].
- My idea: [IDEA].
TASK
Find the specific first customer to win.
DELIVERABLES
1. The narrow segment with the most acute version of the problem.
2. Why this segment is the best place to start (pain, reachability, willingness to pay).
3. The signals that identify this customer.
4. Who I should explicitly ignore for now, and why that is safe.
CONSTRAINTS
- Narrow aggressively; "everyone" is not a beachhead.
- Choose the segment I can actually reach and win, not the biggest one.
- Justify the exclusion of everyone else.
7. Willingness-to-pay probe
This prompt builds a way to test whether customers will actually pay, framing willingness-to-pay as a hypothesis to validate rather than a number to invent. Interest is cheap; payment is the only real signal. It designs the test, it does not fabricate the answer.
You are a pricing analyst designing a test of real willingness to pay.
CONTEXT
- The product: [IDEA].
- The customer: [CUSTOMER].
TASK
Help me find out if people will actually pay.
DELIVERABLES
1. Why stated interest is not evidence of willingness to pay.
2. The ways to test real payment intent, from pre-orders to letters of intent to a paid pilot.
3. The price points to test and how to frame them as hypotheses.
4. What result would count as genuine validation versus polite interest.
CONSTRAINTS
- Do not invent willingness-to-pay figures; design tests that reveal them.
- Treat actual payment or commitment as the only strong signal.
- Be honest that surveys and "would you pay" questions are weak evidence.
8. Alternatives map
This prompt maps everything the customer could do instead of buying from you, including the powerful option of doing nothing. The real competitor is rarely another startup. Understanding the true alternative tells you what you actually have to beat.
You are a competitive analyst mapping the customer's real alternatives.
CONTEXT
- The problem and my solution: [PROBLEM], [IDEA].
- The customer: [CUSTOMER].
TASK
Map what the customer does instead of buying from me.
DELIVERABLES
1. The full set of alternatives: direct competitors, indirect ones, manual workarounds, and doing nothing.
2. What the customer currently uses and why it is good enough or not.
3. What I would have to be clearly better at to make them switch.
4. The switching costs and inertia working against me.
CONSTRAINTS
- Treat "doing nothing" and status-quo habits as the main competitor.
- Do not invent competitor details; mark what I should verify.
- Be honest about the strength of inertia.
Stage 3 — Build the smallest test
The goal now is evidence, not a product. These four prompts help you build the smallest possible thing that tests your riskiest assumption with real people.
9. MVP scoper
This prompt strips your product idea down to the smallest version that tests the riskiest assumption, fighting the founder urge to build everything first. The MVP is an experiment, not a small version of the dream. It defines what to leave out.
You are a product lead scoping the smallest viable test.
CONTEXT
- The full product vision: [IDEA].
- The riskiest assumption to test: [GOAL].
TASK
Define the minimum thing that tests the riskiest assumption.
DELIVERABLES
1. The single assumption this MVP should test.
2. The smallest build that genuinely tests it - possibly not software at all.
3. What to deliberately leave out, and why leaving it out is safe for now.
4. The result that would validate or invalidate the assumption.
CONSTRAINTS
- Scope to test the assumption, not to impress anyone.
- Prefer the cheapest test - concierge, manual, or no-code - over building.
- Resist adding features; every addition delays the learning.
10. Experiment designer
This prompt turns a vague "let's see if this works" into a falsifiable experiment with a clear success threshold defined in advance. Without a pre-set bar, founders rationalize any result as encouraging. It forces you to decide what would change your mind.
You are a growth scientist designing a falsifiable experiment.
CONTEXT
- What I want to learn: [GOAL].
- My situation: [STAGE].
TASK
Design an experiment with a clear pass/fail bar.
DELIVERABLES
1. The hypothesis, stated so it can be proven wrong.
2. The test design and what I will measure.
3. The success threshold, set before running, that separates a real signal from noise.
4. What I will do if it passes, and what I will do if it fails.
CONSTRAINTS
- Set the success bar in advance so I cannot rationalize the result.
- Make the hypothesis genuinely falsifiable.
- Keep the test small and fast.
11. Positioning for the test
This prompt writes the positioning and landing message for your experiment so the test measures real demand, not confusing copy. If people do not understand the offer, a failed test tells you nothing. It makes the value proposition legible enough to actually test.
You are a positioning expert writing the message for my test.
CONTEXT
- The product and who it is for: [IDEA], [CUSTOMER].
- The core value: [PROBLEM] solved.
TASK
Write positioning sharp enough to test real demand.
DELIVERABLES
1. A one-line value proposition that names the outcome and the customer.
2. The headline and supporting message for a test landing page.
3. The single clearest call to action for measuring intent.
4. The objection most likely to stop a sign-up, and how to address it.
CONSTRAINTS
- Specific enough that a competitor could not paste their name into it.
- No hype; the test should measure demand for the real thing.
- Make the value obvious in seconds.
12. Pricing-test framework
This prompt designs a way to test pricing as part of your experiment, treating price as a variable to learn rather than a number to guess. Charging from the start is itself a validation signal. It frames pricing as a hypothesis, not a fabrication.
You are a pricing strategist designing a pricing test.
CONTEXT
- The product: [IDEA].
- The customer: [CUSTOMER].
TASK
Design a test for what people will pay.
DELIVERABLES
1. A starting price hypothesis, framed by the value delivered, not by cost.
2. The pricing model to test (one-off, subscription, usage) and why.
3. How to test price without a large audience - direct offers, pilots, pre-sales.
4. What response would validate the price versus signal it is wrong.
CONSTRAINTS
- Anchor on value, not on cost-plus or undercutting.
- Treat any price as a hypothesis to test, not a fact.
- Charging early is a feature of the test, not a risk to avoid.
Stage 4 — Make the case
If you are raising, the bar is credibility. These four prompts build an honest investor case — and refuse to fabricate the numbers that would blow it up.
13. Investor narrative
This prompt structures the logic of your pitch — the shift, the opening, why you — without inventing traction or market figures. Investors fund a believable story backed by real evidence, not a deck of made-up numbers. It builds the argument; you supply the proof.
You are a pitch advisor building the logic of my investor narrative.
CONTEXT
- The idea and traction so far: [IDEA], [STAGE].
- What I am raising for: [GOAL].
TASK
Structure a credible investor narrative.
DELIVERABLES
1. The one-line version: what we do, for whom, and why it matters now.
2. The three-part story: the shift happening, the opening it creates, why we win.
3. Where my real traction and evidence slot in to support the story.
4. The strongest objection an investor will raise, and the honest answer.
CONSTRAINTS
- Use only my real traction; never invent metrics, customers, or growth.
- Mark clearly where real evidence is needed rather than fabricating it.
- No hype; the narrative must survive a skeptical partner meeting.
14. Investor-objection pre-mortem
This prompt anticipates the questions that kill a raise so you can answer them before they are asked. The objection you did not prepare for is the one that ends the meeting. It surfaces the hard questions while you can still address them.
You are a skeptical investor pre-morteming my raise.
CONTEXT
- The pitch and stage: [IDEA], [STAGE].
TASK
Find the questions most likely to kill the raise.
DELIVERABLES
1. The hardest questions a sharp investor will ask, ranked by how damaging.
2. For each, the honest answer - or the work I need to do before I can answer it.
3. The single weakness most likely to end a meeting.
4. What I should shore up before pitching.
CONSTRAINTS
- Ask the questions a tough investor actually asks, not softballs.
- Do not hand me dishonest answers; flag where I need real evidence.
- Be blunt about the weakest part of the story.
15. Model-assumption auditor
This prompt audits the assumptions inside your financial model, labeling each as sourced or assumed so you never present a guess as a fact. A model built on invented inputs is worse than no model. It exposes the load-bearing assumptions and the ones that would not survive scrutiny.
You are a finance lead auditing the assumptions in my model.
CONTEXT
- My model's key drivers and assumptions: [INPUT].
TASK
Audit the assumptions, not the arithmetic.
DELIVERABLES
1. Each key assumption, labeled as sourced, reasonable, or optimistic.
2. The assumption the whole model is most sensitive to.
3. Which assumptions an investor will challenge hardest.
4. What I should source or test before the model is credible.
CONSTRAINTS
- Do not invent inputs; flag every number that needs a real source.
- Identify the load-bearing assumptions, not just list them.
- Be honest about which assumptions look like wishful thinking.
16. Raise-or-bootstrap framework
This prompt frames the decision of whether to raise capital at all, rather than assuming the answer is yes. Venture funding fits a specific kind of business and harms others. It lays out the tradeoff against your actual goals.
You are an advisor framing the raise-versus-bootstrap decision.
CONTEXT
- The business and its economics: [IDEA], [STAGE].
- What I want from it: [GOAL].
TASK
Help me decide whether to raise.
DELIVERABLES
1. Whether this business fits the venture model (large market, fast growth, scalable) or not.
2. The honest tradeoffs of raising versus bootstrapping for my specific goals.
3. What raising would commit me to that bootstrapping would not.
4. A recommendation, with the factor that should decide it.
CONSTRAINTS
- Do not assume raising is the goal; many good businesses should not.
- Tie the recommendation to my stated goals, not to status.
- Be clear about what taking outside capital obligates me to.
Stage 5 — Grow
Once something is working, the question is where to push. These two prompts focus growth effort where it will actually compound.
17. Growth-channel prioritizer
This prompt identifies which acquisition channel to test first, given your specific customer and economics, rather than spreading thin across all of them. Most early startups win through one channel, not ten. It picks the one to prove before scaling.
You are a growth strategist prioritizing acquisition channels.
CONTEXT
- The product, customer, and economics: [IDEA], [CUSTOMER], [STAGE].
TASK
Decide which channel to test first.
DELIVERABLES
1. The channels that plausibly fit my customer and price point.
2. The one or two most worth testing first, and why.
3. The channels to ignore for now, and why they are wrong for this stage.
4. The cheapest test of the top channel and the metric that proves it works.
CONSTRAINTS
- Concentrate on one channel to prove before scaling; do not spread thin.
- Match the channel to the customer and unit economics, not to what is trendy.
- Define the success metric before testing.
18. Founder focus
This prompt cuts a founder's overwhelming task list down to the few things that actually move the company this week, and what to deliberately ignore. Early-stage focus is the scarcest resource there is. It forces a decision about what not to do.
You are a chief of staff helping a founder focus.
CONTEXT
- Everything on my plate: [INPUT].
- The one thing the company most needs right now: [GOAL].
TASK
Cut this to what actually matters.
DELIVERABLES
1. The two or three things that genuinely move the company this week.
2. The one task that, if done, makes others easier or unnecessary.
3. What to drop, defer, or delegate - explicitly.
4. The busywork that feels like progress but is not.
CONSTRAINTS
- Force real cuts; a founder who does everything does nothing well.
- Tie priorities to the company's biggest current risk, not to what is comfortable.
- Be willing to tell me to stop doing something.
The founder stack: running them as one workflow
These prompts compound in order. Try to kill the idea and validate the problem, sharpen the customer and the demand, build the smallest test of your riskiest assumption, make the honest case if you are raising, then concentrate growth. The thread is that the model's job is to find what is wrong and design what would prove it, while you go get the real evidence from real customers. For the deeper strategic frameworks once you have traction, the business strategy prompt library goes further, and the patterns behind every prompt here live in the prompt library pillar.
The Bottom Line
The worst thing AI can do for a founder is agree with them, and by default that is exactly what it does. The best thing it can do is play the role most founders lack — the smart skeptic who tries to kill the idea, names the assumption you have been avoiding, and insists you test before you build. That is uncomfortable and it is the entire point, because the market will be far less gentle than a prompt. The eighteen prompts here turn the model into that skeptic and that experiment designer. Use it to find the truth about your idea faster than your competitors find it about theirs — and remember that the only real validation comes from customers, not from a model that wants you to feel good.
Can AI validate my startup idea?+
AI cannot validate an idea — only real customers and real evidence can. But it is excellent at pressure-testing your thinking: arguing the case against your idea, surfacing the assumptions it depends on, and designing the experiments that would produce real validation. Treat it as a sparring partner, not a verdict.
What is the biggest mistake founders make using AI?+
Asking it whether their idea is good. Models default to being agreeable and will generate a confident, encouraging analysis of almost anything, which feels like validation and is not. The fix is to explicitly prompt it to attack the idea and to design tests against real customers.
Should I trust AI for market size and financial projections?+
Not the numbers themselves. A model will produce precise-sounding market sizes, growth rates, and financial figures that are fabricated. Use it to build the structure of a model and to label every assumption, then source the real inputs yourself — especially for anything you will put in front of investors.
How do I use AI for customer discovery?+
Use it to design unbiased interview questions and to analyze patterns in the answers you collect, not to invent what customers think. The model can help you avoid leading questions and spot themes across interviews, but the signal has to come from real conversations with real potential customers.
How should I sequence these prompts?+
Start by trying to kill the idea and validate the underlying problem, then sharpen who the customer is, design the smallest experiment that tests your riskiest assumption, build the honest investor case if you are raising, and prioritize a growth channel. Each stage feeds the next and moves you from guessing to evidence.