Asking people if they'd buy it is not validation

Building is no longer the hard part. AI can turn a rough idea into a 60-page guide, a landing page and a sales email before lunch, which means the old filter is gone. It used to take you three months to build something, and somewhere in month one you'd realise nobody wanted it. That delay was doing you a favour. Now there's nothing between the idea and the finished product except you, and most people skip the only step that decides whether it sells. I did too, and the fix isn't a survey.

The version of this I got wrong

I've spent the last few years building with AI, and the pattern I fell into early looks like this. Idea arrives on a Sunday. It feels obvious. By Wednesday there's an outline, by Friday there's a draft, and by the following week there's a product with a cover, a price and a checkout link.

Then nothing. Not rejection, which at least tells you something. Just silence. No sales, no feedback, no signal at all, because nobody was ever waiting for it.

And here's the uncomfortable part: the product was fine. The writing was fine. The design was fine. It failed for a reason that had already been decided before I opened a single tool.

"That sounds great, I'd buy that"

This is the line that ruins more first products than bad writing ever has.

You describe your idea to your partner, your group chat, a Facebook group, maybe a poll on your story. People say it sounds useful. Someone says they'd definitely buy it. You take that as a green light and go and build.

The problem is that saying yes costs them absolutely nothing. It's the polite answer. It ends the conversation, it makes them feel supportive, and it commits them to precisely zero. Rob Fitzpatrick built a whole book around this failure mode, The Mom Test, and the short version is that people will lie to you rather than be unkind to you.

Every piece of advice on this topic tells you to run a survey, put up a landing page, start a waitlist. Fine tactics, except almost all of them assume you already have an audience or an email list to point at them. If you're starting out you have neither. That's the actual position most people are in, and it's the one nobody writes for.

Validation is evidence that costs someone something

This is the shift, and it's the whole thing.

Stop trying to collect opinions. Start trying to collect evidence that cost a stranger something. There are exactly three currencies that count, and you don't need an audience to get any of them.

1. Their words, in their language

Not "would you buy this". Instead, find where your kind of person already complains in public, and read what they typed when nobody was selling to them. Reddit threads, Facebook group posts, YouTube comments under videos on your topic, Amazon reviews of books adjacent to your idea (especially the two and three star ones, where people say what was missing).

What you're hunting for is repetition. The same frustration, phrased in similar words, by different people who don't know each other. That repetition is a real problem with a real shape. One person's complaint is noise; eleven people describing the same wall is a product.

Here's where AI actually earns its place. Paste 30 or 40 of those raw posts in and ask it to group them by the underlying problem, keeping the original wording. Don't ask it to summarise. Ask it to cluster and quote. You want their sentences, not a tidy paraphrase, because their sentences are your sales page later.

This costs the reader nothing, so on its own it's weak evidence. It's the starting map, not the proof.

2. Their time

Time is the first currency with a real price tag on it.

Go back to those same threads and be useful in public. Answer the question properly, for free, in detail. No link, no pitch, nothing to click. Then watch what happens.

If people reply, ask follow-up questions, DM you, or say "do you have this written down anywhere", they've spent something. Attention and effort are finite and they just gave you some. If your genuinely helpful answer lands with a thud and gets ignored, that is a data point too, and it cost you an afternoon instead of a fortnight.

The strongest version of this signal is unprompted. Someone asks you for more without you offering it. That's a person who has already decided the problem is worth solving, which is the thing you were trying to find out.

3. Their money

This is the only signal that settles the argument, and you can get it before the product exists.

Sell the outcome, not the file. Offer to solve the exact problem for one person, manually, this week, for a small price. If your idea was a guide on making short-form videos for local businesses, don't write the guide. Make three videos for one local business for a modest fee. If your idea was a course on AI-written property listings, write ten listings for one estate agent.

Do that two or three times and you'll have three things you cannot get any other way: proof people will pay, the actual step-by-step process you used (which is now your product outline, already tested), and the specific questions they asked you at every stage, which are the exact sections your product needs.

That's also the answer to the 2026 problem with AI-made products. Buyers can spot a fully generated guide instantly: generic examples, no point of view, nothing that couldn't be prompted in five minutes. The products that sell are AI-assisted and human-led, and the human part is precisely what you learned doing the work by hand first. Validation and differentiation turn out to be the same step. This is the logic the whole Practical Income Method runs on: demand first, delivery second, product third.

The rule I use before I open a single tool

One week of looking, then a decision.

Day one to three: read and cluster. Find the repeated complaint in the words people actually used. If you can't find repetition, the idea isn't ready and no amount of building will fix that.

Day four to five: answer in public, for free, at least five times. Watch for pull.

Day six to seven: offer to do it once, by hand, for money. Even £20. The number matters far less than the fact that somebody reached for their card.

If nobody pays, you haven't lost a product. You've lost a week and gained the reason it wouldn't have sold, which is worth considerably more than a finished file sitting in a folder.

This works precisely because you have no audience

If you had 10,000 followers you'd be tempted to poll them, get 400 likes and build the wrong thing with confidence. Starting with nobody forces you into public places where people are already stating their problems out loud, and forces you to earn the first yes from a stranger. That is a harder test and a better one.

You don't need a list. You don't need a platform. You need one stranger to hand over one payment for one small version of the thing, and you can go looking for that this week.

Validation is not applause. It is evidence that cost somebody something.

If you want the full sequence, from finding the problem to shipping the product to selling it without an audience, I've put the whole thing into a free five-step guide: the Practical Income Method. It's the exact order I'd follow if I were starting again on Monday, with the validation week built in at the front where it belongs.