Validating An Idea With AI (Without Fooling Yourself)

Validation means: before you build, find out if anyone actually wants this. AI has made validation 10x faster. It has also made it easier to skip — to ask Claude 'is this a good idea' and walk away with a false thumbs-up. Validation has to involve real humans with wallets. AI helps you get to those humans faster.

Move 1: 10 customer conversations

The single highest-value thing you can do before writing code is have 10 conversations with potential customers. Not 'do you like my idea' conversations — those produce lies. 'Tell me about the last time you had this problem' conversations. Those produce truth.

An AI-assisted interview script

Ask Claude to generate an interview script for you:

"I'm validating this idea: [one-paragraph description]. My customer is [description]. Write a 30-minute interview script with 8-10 open questions that probe their past behavior around this problem. Every question should get at how they currently solve it, not what they'd do in the future. Do not put a single 'would you' question in the script."

Then run the interviews. Record them (with permission). Afterward:

"Here's a transcript of my interview with [customer]. Identify:
1. The strongest 3 direct quotes about the problem.
2. What they are currently paying for or doing (not what they said they'd do).
3. Any signals that my proposed solution would actually fit.
4. Any signals I'm hearing what I want to hear."

Move 2: the smoke-test landing page

Build a landing page before you build the product. Clear headline, clear promise, one CTA. Run $100 of targeted ads or post it in 5 communities where the customer hangs out. Count the click-to-signup rate. If zero people sign up, you don't have demand. If 20% sign up, you might. Build the product next.

• Tool: v0.dev — describe the page, get working code, deploy to Vercel in one click
• Headline: 1 sentence that names the customer and the outcome
• Proof: social proof or founder story, kept short
• CTA: email capture, or 'pay $1 to reserve' if you're bold

Move 3: the 'I will build this if' preorder

The ultimate validation: ask for money before the product exists. A preorder, a waitlist deposit, a 'commit to buy' pledge. If people hand you real money or write real emails with intent, the idea has blood. If not, find a new idea.

Using AI to find your first 10 interview subjects

"My target customer is [specific description — e.g. small-practice dentists in the US with 1-3 locations]. Help me find 10 of them to interview. For each suggestion, give me:

1. Where they hang out online (specific subreddits, Facebook groups, LinkedIn communities, Slack groups, forums).
2. A 2-sentence cold DM I could send them that respects their time.
3. A small 'gift' I could offer in exchange for 20 minutes (a free audit, a custom resource, etc).
4. One reason they might say no and how to address it."

The self-deception alarm bells

• You've done more design than interviews — warning sign
• Every interview 'confirmed' your thesis — you're leading the witness
• You're defending the idea instead of interrogating it — ego is driving
• You've delayed interviews for more than 2 weeks — you're avoiding truth

What 'good' looks like

A good validated idea has: 10 real interviews documented, 3+ people who said 'when is it ready, I'll buy,' at least one signal of willingness to pay, and a landing page converting above 10% from targeted traffic. If you have those four things, build. If not, don't — keep validating. Building the wrong thing is the most expensive thing you can do.