How to Test Your Product When You Have No Users

You can test your product with no users by running it past three sources: people in your network, paid participants from a recruiting panel, and AI testers that drive your app in a real browser. The cheapest path that scales is AI testers — tools like Cast spin up personas that actually click through your app, record the session, and tell you what's broken, so you find the obvious friction before a real user ever hits it.

The hard part isn't analyzing feedback. It's getting any feedback at all when nobody's signed up yet. Here's how to do it without a budget, a panel, or a waitlist.

Why "no users" is the wrong reason to skip testing

Testing without users is possible because usability problems show up the same way no matter who finds them. A confusing signup, a button that does nothing, a form that loses your input — these break for the first tester exactly like they break for the thousandth.

Nielsen Norman Group's research is the reason this works: testing with just five users surfaces the majority of usability problems in a flow. You don't need a crowd. You need a handful of people — or stand-ins — who try to finish a real task and tell you where they got stuck.

The catch is finding even those five. That's where the cost shows up.

The real cost of recruiting testers

Recruiting is the expensive, slow part — not the testing itself. According to Nielsen Norman Group, the average cost to recruit one usability participant is $171, and internal recruiting eats about 1.15 hours of staff time per person. Roughly one in nine people who agree to show up don't (an ~11% no-show rate, which 2026 platform data puts closer to 20–30%).

The downstream numbers are worse. According to User Intuition's 2026 budget guide, a fully-loaded 10-participant moderated study runs $8,000 to $20,000 once you count researcher time. CleverX pegs the median cost of a single user-research study in 2026 at $3,200. And per Vendr's contract data cited by Articos, the average SMB pays $40,128 a year for UserTesting.

For a solo builder pre-launch, those numbers are the whole problem. So the practical move is to start with the methods that cost little or nothing.

Five ways to test with no users

Here are the methods that work before you have customers, ranked roughly from cheapest to most thorough.

1. Guerrilla testing. Ask five people — friends, a coworking neighbor, your Discord — to do one task while you watch silently. Cost: coffee. Best for catching the most obvious breaks early.
2. Build-in-public feedback. Post a specific ask ("what's confusing on this signup?") to Reddit, Indie Hackers, or X. You'll get noise, but the repeated phrases are signal.
3. Unmoderated panels. Tools like Maze or Lyssna route paid participants through a task. Fast, but you still recruit and pay per head — $30–$150 a person.
4. AI agent-based testing. AI personas open your app in a real browser, pursue a goal, and record what happens. No recruiting, no scheduling, results in minutes. Best for continuous, pre-user testing.
5. Moderated interviews. A live call where you watch someone work. The richest signal — and the most expensive in time once you have users worth interviewing.

How the methods compare

Where AI testers fit — and where they don't

AI testers are strongest at behavior and weakest at judgment, so use them for what they actually do. An AI persona that drives a real browser is genuinely good at the mechanical truth: did the button work, did the flow complete, where did it dead-end. Amazon Science's 2026 UXAgent framework showed LLM agents can generate thousands of simulated sessions with video recordings for exactly this kind of evaluation.

Be honest about the limits. A 2026 review summarized by Koji found that synthetic survey personas "hallucinate insights" and are "individually believable but collectively wrong" — as Pavel Samsonov put it, "an LLM can't buy your product." That critique is real, and it's aimed at AI that answers questions as if it were a customer. It does not describe an AI that navigates your app and records what breaks.

The distinction matters. Nielsen Norman Group cautions that AI "is not actually watching" where a user looks — so the answer is to keep the receipts. A tester that records the full session as video gives you something a human can review, not a confident summary you have to trust blind.

So: use AI testers for usability and bugs. Use real humans — eventually — for willingness to pay and the unexpected.

How Cast tests your product before you have users

Cast spins up AI users who actually test your app, record the session, and tell you what's broken. Each persona gets a believable backstory, a real frustration, and one concrete goal — then it interacts with your product the way a real user would and writes honest feedback in its own voice: usability notes, bugs it hit, and product ideas.

There's no one to recruit and no calendar to chase. Drop in your URL and the first session comes back in minutes, recorded so you can watch exactly where the persona got stuck. Personas remember across runs, so the feedback compounds instead of resetting. The free tier includes five personas, no credit card, and the Builder plan is $39/month.

When you have no users, your next tester is one URL away. Create your first persona →

FAQ

Can you really test a product with zero users? Yes. Usability problems break the same way for any tester, so guerrilla sessions, paid panels, and AI testers all surface real issues before you have customers. Nielsen Norman Group's research shows five testers find most usability problems in a flow.

Are AI user testers accurate? For behavior — clicking through flows, hitting bugs, finding dead ends — they're reliable, and a recorded session lets you verify what happened. For judgment calls like pricing or "would you buy this," validate with real people. AI is strong on what broke, weak on what someone would pay.

How much does user testing cost in 2026? The median traditional study costs about $3,200 according to CleverX, and recruiting alone averages $171 per participant per Nielsen Norman Group. AI-based testing starts far lower — Cast's Builder plan is $39/month with no per-session recruiting cost.

How many testers do I actually need? Around five per flow catches the majority of usability problems. The constraint is rarely the number — it's the time and cost of finding them, which is why builders with no users start with AI testers.

What's the fastest way to get feedback before launch? Run your live URL past an AI tester. There's nothing to schedule, results come back in minutes, and you get a recorded session plus written feedback — fast enough to fix issues the same day you find them.