Synthetic Users, Used Honestly
You've seen the breathless claims that AI participants will replace real users. We don't believe that either. Synthetic users are a complement with specific, defensible uses: piloting discussion guides before you spend real participants on them, exploring hypothesis space cheaply, stress-testing edge-case perspectives you can't recruit, and giving stakeholders a visceral preview of research value. Real findings still come from real people. The question is how much sharper you arrive.
The research gap you're working around
Real participants are too expensive to waste
A flawed discussion guide discovered mid-study burns recruited, incentivized, scheduled humans. Piloting on synthetic participants costs minutes.
Coverage never matches demand
Five product teams, one researcher. The decisions that don't get research still get made, just without you.
Edge perspectives resist recruiting
The churned enterprise admin, the accessibility-dependent power user, the procurement gatekeeper: real ones are nearly unrecruitable on a deadline.
What you can run, starting today
Pilot the discussion guide
Run your interview protocol through an AI-facilitated session first. Leading questions, dead ends, and ordering problems surface before a real participant ever sees them. AI Deep Dive.
Explore before you commit
Ten cheap directional probes across hypothesis space, then spend your real-participant budget on the two that matter. AI Focus Group.
Build the unrecruitable panel
Generate personas for perspectives you can't schedule: specific roles, constraints, and attitudes, reusable across studies. Persona Generator.
Give stakeholders the demo
Show a skeptical PM what a focus group does by running one in front of them. It makes the case for real research better than any slide. AI Personas.
An illustrative example
A solo researcher at a fintech piloted a sensitive interview guide about financial stress before a real diary study.
Three deep-dive runs with personas across financial situations, each testing a revision of the guide.
- The opening question ('tell me about your budget') shut the first persona down; reordering to start with routines, not numbers, opened the conversation.
- A question the team considered essential turned out to be answerable only after trust-building two questions later: ordering, not wording, was the fix.
- By run three the guide produced consistently deep responses, and the revisions had cost zero real participants.
The decision: The real diary study launched with a third-draft guide, and the first human interview produced usable depth instead of being the rehearsal.
Built for the way you work
- Pilot runs cost minutes, so guides arrive at real sessions pre-debugged
- Directional exploration scales to every team that asks, not just the loudest
- Synthetic results are labeled synthetic, always: no laundering them into findings
- Transcripts and reports export for your repository like any other study
Frequently Asked Questions
Are synthetic users valid for UX research?
For generating hypotheses, piloting instruments, and directional exploration: yes, with eyes open. For validated findings about real human behavior: no. The literature is clear that LLM-based participants exhibit real biases and miss embodied context. Use them where being roughly right fast matters, and real participants where being actually right matters.
Won't this encourage teams to skip real research?
It's designed to cut the other way: a team that has seen a panel disagree with them tends to get curious what real users would say. That's why results are explicitly labeled directional and synthetic, with limits stated on the report.
What's the strongest use case for a research team?
Guide piloting. It's pure upside: every flaw caught synthetically is a real participant saved, and the method risk is near zero because no finding is claimed from the pilot itself.
Can I cite synthetic runs in a research report?
Cite them as what they are: synthetic pilots or exploration that shaped the study design. Including a short 'how we used synthetic participants' methods note reads as rigor, not corner-cutting.
Keep exploring
Put a panel to work today
Start free and run your first study in minutes. Get started.