What are synthetic users? A practical guide for founders

Every founder knows they should talk to users before they build. Most don't do it often enough, and not because they don't care. Recruiting the right people, scheduling calls and turning hours of notes into decisions takes weeks. By the time the answers arrive, the page has already shipped.

Synthetic users are one answer to that gap. This guide explains what they are, what they're genuinely useful for, and where you still need real people.

What is a synthetic user?

A synthetic user is an AI-generated stand-in for a type of customer. Instead of a one-page persona document that sits in a slide deck, a synthetic user has a detailed profile (a role, a budget, habits, goals, objections, patience) and can respond to things in character.

You can ask a synthetic user a question, show it your landing page, or hand it a pricing table, and it will tell you how someone like them would react and why.

The words get used loosely. You'll also see AI personas, simulated users or digital customers. They all describe the same idea: a language model prompted with a rich, consistent description of a person, used to predict how that kind of person would respond.

What synthetic users are good at

Synthetic users are strongest where the problem is visible on the page and the reasoning is common sense for the person you're simulating.

• Spotting unclear messaging. If your headline only makes sense to people who already know your product, a persona who doesn't will say so.
• Surfacing objections. A cautious IT buyer asks about security reviews. A deal-savvy shopper asks about discount codes and returns. These are predictable, and predictable is exactly what you want to catch before launch.
• Finding missing trust signals. No pricing, no customer names, no clear next step: a sceptical visitor notices immediately.
• Comparing audiences. The same page can land well with developers and badly with their managers. Seeing both side by side tells you who the page is really written for.
• Fast iteration. Because a study takes minutes, you can change the headline and run it again in the same afternoon.

Where they fall short

Being honest about the limits is what makes synthetic users useful rather than dangerous.

• They can't tell you what people will actually pay. Stated willingness to pay is unreliable even from real people. From a simulation, treat it as a prompt for questions, not a number.
• They don't know your market's private context. If your buyers all read the same niche forum or follow a specific regulation, a persona only knows what you tell it.
• They regress to the plausible. Simulations are good at typical reactions and weak at the surprising ones, the odd workflow or unexpected use case that real research uncovers.
• They can't replace usage data. What people say and what they do are different things. Analytics and real behaviour are still the final word.

How to use synthetic users well

The best way to think about synthetic users is as the first round of research, not the last. Use them to remove the obvious problems so that the time you spend with real customers goes on the questions that only real customers can answer.

A simple workflow:

1. Describe your audience precisely. "Small business owners" is too broad. "Café owners with one location who do their own marketing between shifts" gives a persona something to reason from.
2. Test the real artefact. Show personas the actual landing page or post, not a description of it. Reactions to the real thing are more specific.
3. Read the reasons, not just the scores. A low trust score tells you something is wrong. The quote explaining it tells you what to change.
4. Fix, then re-run. Change one thing at a time and test again to see whether it helped.
5. Take the open questions to real people. Whatever the personas disagreed on, or whatever you still can't decide, is what your customer calls should focus on.

A quick example

Say you're launching a scheduling tool for clinics. You run a study on your homepage with three personas: a time-poor GP, a practice manager and a clinic owner.

The GP skims, likes the promise of fewer admin hours, but leaves because she can't tell whether it works with her existing records system. The practice manager wants to know who handles patient data. The owner wants a price.

None of these is a surprise, and that's the point. In ten minutes you've found three fixes (an integrations line above the fold, a data-handling section, a visible pricing link) that would otherwise have cost you your first week of traffic.

The bottom line

Synthetic users don't replace talking to customers. They make the conversations you do have more valuable, by catching the predictable problems first. If you're launching something soon, they're the cheapest feedback you'll get.

If you want to try it, betaproof builds AI personas of your audience from a link or a short brief, and runs studies on your pages and posts in minutes. You can see what a finished report looks like in the sample study.