The ChatGPT prompts for startups below help founders do the work that decides whether a company survives: testing the idea, sizing the market honestly, scoping a first version, finding early users, modeling runway, and talking to investors. Each prompt asks the model to push back and show its reasoning, because an AI that agrees with every idea is no use to a founder.
How to use these prompts
Copy a prompt into ChatGPT, Claude or Gemini. Replace the blanks in double braces, or keep the default after the colon: {{stage:pre-revenue}} means "pre-revenue" unless you change it.
Give the model the real material: your customer interview notes, your actual numbers, the investor's email. A model with no context produces the startup advice found in every blog post. Keep each prompt to one job and run them in sequence. OpenAI's ChatGPT prompting guide puts it simply: "Tight scope usually beats a 'do everything' prompt."
Treat every number the model gives you as a draft. Market sizes, benchmarks and growth rates from a model can sound exact without a source. The prompts below ask for math built from inputs you supply, so you can check every step.
Which prompts test the idea?
1. Find the riskiest assumptions
Start here. Every startup idea rests on a few beliefs that, if wrong, end the company.
Here's my startup idea:
{{idea}}
Target customer: {{customer}}
Stage: {{stage:pre-revenue}}
Act as a skeptical, experienced founder. List the assumptions this idea
depends on, across these areas: the problem is real and painful, this
customer has it, they'll pay, we can reach them affordably, we can build it,
and nobody can easily copy it.
Rank the assumptions by how likely they are to be wrong times how fatal
that would be. For the top three, say what evidence would prove each one
wrong.
2. Design cheap tests for each assumption
Turns the list from prompt 1 into experiments you can run this month, most of them without writing code.
For each assumption below, design the cheapest test that would give real
evidence within {{timeframe:two weeks}}. Options include customer calls, a
landing page with a signup, a concierge version done by hand, a pre-sale,
or a manual pilot.
For each test: what we do, how many people we need, what result would
count as a pass or a fail (decide this before running it), and the cost in
time and money. Prefer tests where people act (sign up, pay, give time)
over tests where they only say they like it.
Assumptions:
{{assumptions}}
3. Bottom-up market size
Top-down numbers ("1% of a $50B market") are easy to write and hard to trust. This prompt builds the estimate from counts and prices.
Help me estimate the market for {{product}} bottom-up.
Inputs I have:
- Who buys: {{buyer:operations leads at US logistics companies with 50 to 500 employees}}
- How I'd count them: {{count_source}}
- Expected price: {{price:$300 per month}}
Build the estimate step by step: number of potential buyers, the share that
realistically has the problem, the share we could reach in three years, and
revenue at our price. Show every multiplication. Label each input as
"mine" (I gave it), "assumption" (you proposed it) or "needs a source".
Don't use any outside statistic without marking it "needs a source".
Finish with a low, middle and high case.
4. Map the alternatives customers use today
Your real competition is often a spreadsheet, an intern or doing nothing.
For {{customer}} trying to solve {{problem}}, list the alternatives they use
today: direct competitors, indirect tools, manual workarounds (spreadsheets,
email, people), and doing nothing.
For each: why someone picks it, what it costs them (money and time), and
what would make them switch. Then tell me which alternative is hardest to
beat and why. Base competitor details only on what I paste below; if I
haven't pasted anything about a competitor, say what I should look up.
What I know about competitors:
{{competitor_notes}}
Which prompts help you build and launch?
5. Scope the smallest useful version
Founders overbuild. This prompt makes the model argue for cutting.
Here's the feature list for the first version of {{product}}:
{{features}}
The one job the first customers hire it for: {{job}}
Sort every feature into: must have for that job, can be done by hand at
first, can wait, and cut. Be aggressive. For each "done by hand" item, say
how we'd fake it manually. Then describe the first version in three
sentences and estimate what we drop by building only that.
6. Pricing hypotheses to test
Pricing is a guess until someone pays. This prompt gives you guesses worth testing.
Help me form pricing hypotheses for {{product}}, sold to {{customer}}.
What it replaces and what that costs them today: {{current_cost}}
What our costs per customer look like: {{unit_costs}}
Propose three pricing models (for example per seat, usage-based, flat tier)
with a starting price for each and the reasoning. For each, name the
customer behavior it rewards or punishes. Then suggest how to test them
with the next {{count:10}} sales conversations without confusing prospects.
7. Plan the first 100 users
Early users come from places where your customers already gather, not from launch day traffic.
I need my first {{goal:100}} users for {{product}}. Ideal user:
{{customer}}. My budget: {{budget:almost none}}. My unfair advantages
(network, audience, expertise): {{advantages}}.
List ten specific channels where these people already spend time or look
for help (communities, events, newsletters, marketplaces, partners, direct
outreach). For each: how I'd approach it without spamming, the effort
involved, and how many users it might realistically bring. Rank them, and
give me a plan for the first two weeks using the top three.
Which prompts help with money and investors?
8. Runway and burn from my numbers
Every founder should know their runway without opening a spreadsheet. This prompt shows the formulas so you can rebuild it.
Here are my numbers:
- Cash in the bank: {{cash}}
- Monthly costs by category: {{costs}}
- Monthly revenue and growth rate: {{revenue}}
- Planned hires and their start months: {{hires}}
Calculate monthly net burn and the month cash runs out, under three cases:
as planned, revenue growth at half the rate, and hires delayed three months.
Show the formulas so I can rebuild this in a spreadsheet. Then tell me the
one or two levers that change runway the most.
9. Pitch deck outline
For a first investor deck. The model proposes structure; your evidence fills it.
Outline a pitch deck for {{company}}, raising {{round:a pre-seed round}}.
What we have: {{material}}
Give {{slides:10 to 12}} slides with a conclusion headline for each and the
evidence it needs. Use only facts from my material. Where a slide needs
evidence I don't have yet (traction, team, market), mark it [NEED: ...]
instead of filling it. Then tell me which slide is weakest and what would
fix it.
10. Practice investor questions
Rehearse the hard questions before the meeting, not during it.
Act as an experienced investor in a first meeting with me about
{{company}}. Here's my pitch summary: {{summary}}
Ask me one question at a time, the way a skeptical investor would: market,
why now, competition, traction, team, unit economics, and use of funds.
After each answer, tell me in one or two lines what was weak or vague and
what a stronger answer would include, then ask the next question. After ten
questions, summarize my three weakest areas.
11. Understand a SAFE or term sheet
Y Combinator publishes its standard SAFE documents, and describes a SAFE as "a short contract an investor signs to fund your startup now". This prompt helps you read one before you talk to your lawyer; it is not legal advice.
I'm a first-time founder. Explain the document below in plain language.
1. Summarize what each key term means for me: amount, valuation cap,
discount, pro rata rights, and anything unusual.
2. Show with a simple example how my ownership would change at a future
priced round of {{example_round:$5M at a $20M pre-money valuation}}.
Show the math.
3. Flag any terms that look non-standard compared with a typical SAFE.
4. List the questions I should ask my lawyer.
Say clearly when you're unsure what a clause means. Don't guess.
Document:
{{document}}
12. Monthly investor update
Short, regular updates build trust and make your next raise easier. Include the bad news.
Write my monthly investor update from the notes below. Format:
- One-line summary of the month.
- Key metrics vs last month: {{metrics}}
- Wins (three at most).
- Challenges, stated plainly, with what we're doing about each.
- Specific asks: intros, hires, advice.
Under {{words:300}} words. Use only my numbers and don't spin a miss as a
win.
Notes:
{{notes}}
Which prompts help run the company?
13. Plan the week around one goal
Founders drown in tasks. This prompt sorts your list against the goal that matters this quarter.
This quarter's most important goal: {{goal}}
Here's everything on my list this week:
{{tasks}}
Sort each task into: moves the goal directly, must be done but doesn't move
the goal, can be delegated, can be dropped. Then give me a plan for the week
that protects at least {{focus_hours:10}} hours for goal work, and tell me
which one task I'm most likely avoiding.
14. First hire: job description and interview plan
A first hire shapes the company. This prompt makes you define what success looks like before writing the ad.
I'm hiring my first {{role:full-stack engineer}} at {{company}}, a
{{stage:pre-revenue}} startup building {{product}}.
1. Define what this person must accomplish in their first 90 days.
2. Write a job description built around those outcomes: the work, who
thrives in it, honest trade-offs of joining this early, and what we
offer. No "rockstar" or "ninja" language.
3. Propose an interview process of three or four steps, with what each step
tests and a scoring rubric for the main skill.
How do you stop the model from flattering your idea?
Ask it to argue the other side and to say when it doesn't know. Anthropic's guide to reducing hallucinations suggests explicitly allowing the model to admit uncertainty and limiting it to information from the documents you provide. Prompts 3, 4 and 11 build that in. Two more habits help: run prompt 1 in a fresh chat, so earlier enthusiasm in the conversation doesn't color the answer, and ask "What would make this fail?" after any answer that sounds too good.
What should founders never hand to a model?
Keep decisions that need judgment or a professional with you: legal terms, tax and equity questions, and anything a customer or investor will rely on. Use the model to prepare: summarize, list questions, draft. Before pasting investor documents, customer data or anything under NDA, check your agreements and your app's data settings.
Keep your founder prompts in one library
The idea, pricing and investor prompts get run again every time the plan changes. For pitch deck slides and speaker notes, see ChatGPT prompts for presentations. If you want your coding agent to follow one of these the same way every time, turn it into a skill with the SKILL.md generator.
Kipt keeps prompts like these as Markdown in a private GitHub repo you own, with tags to group them by stage and ⌘K to find the runway or investor update prompt when you need it.