Validate an idea before you build it
Pressure-test the idea so you don't build something nobody wants.
Jervis Koh · tested in my own business
Working realtor, The Real Matters. If a prompt doesn't survive real clients, it isn't here.
Before building anything, have AI play a skeptical customer and tell you why they wouldn't buy. Then test real demand cheaply with a landing page or a handful of real conversations.
The most expensive way to test an idea is to build it; the cheapest is to talk about it. Most failed projects nailed the execution of something nobody wanted. AI gives you a fast, free way to pressure-test an idea — as long as you use it to find the holes, not to feel good about yourself.
- ▸Claude / ChatGPT — role-play a skeptic and map the risky assumptions
- ▸Carrd / Framer — a quick landing page to see if people will sign up
- ▸Google Forms / Tally free — a fast survey to gather real responses
Make AI argue against you
Your own enthusiasm is the enemy of honest validation, so deliberately recruit a skeptic. Ask AI to play a doubtful potential customer and tell you why they wouldn't pay, what's confusing, and what would have to be true for them to use it. The instruction that matters is 'be honest, not encouraging' — left to default, models are agreeable, and a yes-man teaches you nothing. A good skeptical pass surfaces the obvious objections you've been carefully not thinking about.
Stress-test the assumptions
Go beyond a single skeptic and have AI map the assumptions your idea quietly depends on — that people have this problem, that they'd pay to solve it, that they'd choose you over the alternatives (including doing nothing). For each, ask how you could cheaply test it in the real world. AI turns a vague 'I think this is good' into a list of specific, falsifiable beliefs, and the riskiest one is what you test first.
Then talk to actual humans
The firm limit: AI can rehearse objections and structure your thinking, but it cannot confirm demand — only real people parting with real time, money or email addresses can. Treat the AI pass as preparation. Use it to sharpen the questions you'll ask, design a quick test (a landing page, a pre-order, a few conversations), and make sense of the responses. The fatal mistake is mistaking a convincing AI chat for market validation.
- 1Ask AI to play a doubtful customer and list the real reasons they would not pay.
- 2Have it name the single riskiest assumption the whole idea depends on.
- 3Decide in advance what a clear yes looks like (say 20 sign-ups, 5 deposits, or 3 booked calls) so you cannot move the goalposts later.
- 4Design the cheapest possible test of that one assumption: a landing page with a sign-up, a pre-order, or five real conversations.
- 5Build only that test, not the product.
- 6Put it in front of real people and count actions, not compliments; an email, a deposit or a booked call is a signal, praise is not.
- 7Build only once real people take a real action. If they do not, change the idea or the audience and test again.
Idea-validation prompts
Was this useful?
I drop one like this most days in the channel — free, no spam.
Want the prompts that go further?
Drop your email + WhatsApp — I'll send the weekly playbook and the full prompt vault.
Prefer to ask first? Message me on WhatsApp · see plans