Synthetic users allow you to validate ideas quickly, simulating real profiles based on qualitative data to reduce timelines and minimise bias in the design process.
Validating ideas with real users is essential, but it can be slow and costly. At Interactius we use AI-generated synthetic users to test hypotheses with agility, without sacrificing empathy or rigour. These realistic archetypes, built from qualitative patterns, allow us to simulate flows, validate proposals and keep the voice of the customer present throughout the design process.
Contrasting ideas with real users is one of the cornerstones of any human-centred design process. But it can also be costly, slow… and sometimes uncomfortable. Because it means facing the possibility that your idea doesn't work, that nobody understands it, or that it simply doesn't interest anyone.
"Nobody wants to discover that their idea is a 💩 —pardon the expression—. You already have an obstacle within yourself."
Néstor Guerra
At Interactius we are exploring the use of AI-generated synthetic users, built from real patterns identified in qualitative interviews. And yes, they work surprisingly well for testing hypotheses quickly and empathetically, without letting your own assumptions get in the way.
What are synthetic users?
They are not invented characters conjured up in a brainstorming session. They are realistic, dynamic archetypes that represent decisions, emotions and behaviours observed in real people.
Thanks to AI —in my case, I use ChatGPT— these profiles come to life. Not just as documents, but as entities you can interact with.
Here is what they allow you to do:
- Simulate usage scenarios.
- React to texts, flows or design proposals.
- Keep empathy active throughout the design process.
They do not replace real users, nor do they validate without a prior base of qualitative data. But used correctly, they accelerate learning and allow you to iterate with agility and purpose.
How to create synthetic users
This is my 7-step process:
- I organise the qualitative interviews: a minimum of 20 per homogeneous segment. I anonymise and group them by key themes: motivations, frustrations, behaviours…
- I identify patterns: I use ChatGPT to spot similarities between participants.
- I define key dimensions: What differentiates them? Role, context, level of experience, goals…
- I group profiles: I create between 2 and 6 representative clusters.
- I generate the synthetic users: ChatGPT helps me define name, age, context, goals, frustrations… and a representative verbatim quote for each one.
- I refine with the team: we cross-check against the data and validate that the profiles "sound" real.
- I make them interactive: I create a GPT that interprets each profile to simulate interviews, test flows, validate hypotheses or explore ideas.
How synthetic users are used
These synthetic users are particularly useful in the early stages of design:
- Rapid concept validation
- Value proposition
- User story mapping
- User flows
- Wireframes and low-fidelity prototypes
Ultimately, each of these artefacts represents a hypothesis, and being able to test it against these archetypes helps you detect friction points before they cost money.
Moreover, quickly accessing the perspective of a specific segment allows you to iterate without slowing down the development pace. In seconds you can realise that:
- A flow makes no sense for someone with that profile.
- You are projecting your own biases.
- You are falling in love with a solution that doesn't fit.
And that, however uncomfortable, saves you time, energy and frustration.
What about real contact with the user?
Nothing replaces it.
Synthetic users do not substitute interviews, tests or live validation sessions.
But they do help to combat confirmation bias — that natural impulse to only seek out what validates your idea. Having these profiles close by is like having the user in the office every day, reminding you that we design for them, not for ourselves.
And just as real users evolve, our synthetic archetypes must evolve too. That is why I update them with new qualitative samples periodically, depending on the product or market.
AI as an ally for iterating more intelligently
Validating ideas is uncomfortable. It feels like hard work. And as Néstor rightly points out:
"Research is a very bitter path, where you don't see results immediately. But it is the only thing that can prevent you from building something that makes no sense at all."
AI does not eliminate that path, but it does make it more agile.
It allows us to build bridges between research cycles, accelerate learning and, above all, avoid self-deception.
Because as he aptly puts it:
"Validating is not about seeking to be right, it is about seeking the truth."
3 takeaways for your team
✅ Iterate quickly, with focus: you can challenge hypotheses from the very first sketch, without waiting weeks to organise a test.
✅ Reduce the risk of building something that doesn't fit: you confront biases, misunderstandings and ideas that only work in your head much sooner.
✅ Continuous empathy without friction: designing with synthetic users is a practical way to keep the voice of the customer present day to day.
Just how reliable can these profiles be? In the second part of this exploration, we compared 20 synthetic users with a real sample of 150 people and achieved a 93% match in their responses. The results were as surprising as they were promising. You can read the full analysis here: AI in UX Research: 20 synthetic users who think like 150 real ones.
Would you like to try this technique with your team?
If you want to know how to start creating your own synthetic users, or even interact with one of them, we would be delighted to show you how we do it.
