Actual
Simulation
Synthetic personas for researchers
ActualSimulation
Turn real interviews into a living persona you can interview back — voice to voice, grounded in verbatim evidence. One interview becomes a persona you can question and a synthetic panel you can measure, any time.
Run a simulation- Recent simulations
- #0421 — The Cautious Optimizer
- Grounding
- Verbatim evidence only
What it's for
A participant in the room when there isn't one. Pressure-test messaging, explore a segment, and rehearse hard questions against a persona built from real evidence — not a vibe.
Built for UX researchers, strategists, and product teams who need fast, faithful answers between rounds of live research.
How it works
Intake
Upload interview transcripts — or run an AI-moderated, voice-to-voice interview with your participants. Speech is transcribed and mined for evidence automatically.
Source & approve
Actual Simulation proposes enrichment from public signals. You approve every source before it can shape the persona — nothing enters without your sign-off.
Synthesize
A living persona is composed from the evidence: traits, motivations, and a contradiction map — every claim cited back to the verbatim line it came from.
Interview
Talk to the finished persona in real time, voice to voice. Ask the questions you wish you'd asked — it answers in character, never beyond the evidence.
From one evidence base
The same grounded evidence powers two modes — go deep with one voice, or broad across a synthetic audience.
Interview
Talk to a single persona, one to one — voice or text. Qualitative depth: ask follow-ups, probe motivations, and hear it answer in character, never beyond the evidence.
Panel
Measure how a whole synthetic audience reacts to a concept, message, or price. Free-text reactions are mapped to rating distributions you can calibrate against real survey data — directional, fast, and traceable to the evidence.
Why it's different
Voice to voice, both ways
Interview real participants by AI voice during intake, then interview the synthetic persona by voice once it's built. No scheduling, no transcription lag.
Grounded, not guessed
Every statement traces to a verbatim excerpt — open any claim to see its source. The persona can't assert what the evidence doesn't support.
Keeps the contradictions
Real people hold opposing views. The contradiction map surfaces those tensions instead of flattening them into a tidy, misleading average.
Your prompts, not a black box
Tune the system prompts behind the interviewer, the synthesis, and the persona's voice — set them per project, or override per study.
Summon your
first persona
Run a simulationEvery persona claim is backed by verbatim excerpts from your interviews and approved sources.
Projects are scoped to your team. Data stays in your account and is never used for model training.