Synthetic users are AI-generated actors used to rehearse research questions before human fieldwork. MiroFish grounds them in uploaded interviews, reviews, support notes, product briefs, and market evidence, then runs multi-role reaction rounds to surface hypotheses, objections, and validation gaps. The output is not representative customer data.
Live decision rehearsal
Multi-role reaction path
Direct answers
Start with the evidence needed for the decision. Use simulation to expose uncertainty, not to hide it behind generated volume.
Synthetic users are AI-generated actors or profiles that respond to a bounded research scenario; they are not recruited people, measured customers, or a statistical sample.
MiroFish turns uploaded interviews, reviews, support and sales notes, product briefs, and market evidence into an inspectable source packet and actor graph.
Use synthetic users for early research design, concept and message rehearsal, and screener refinement; use real users for lived experience, measurement, and observed behavior.
Research guardrails
Current research supports synthetic users as a directional aid for study design and hypothesis generation while warning against using generated participants as a substitute for human evidence.
An original survey of 150 researchers frames synthetic users as a complement for directional exploration and research design, not a replacement for recruited participants.
Read the sourceThis preprint, currently under review, examines failure modes and reinforces the need to validate model-generated findings against real people and real behavior.
Read the sourceFive-step workflow
Every step leaves something inspectable: the source, the actor assumptions, the reaction path, or the next human check.
State the product, message, or market decision, the intended audience, and the cost of being wrong.
Add interviews, reviews, support themes, sales notes, product context, and relevant market evidence.
Review each actor's grounding, assumptions, relationships, missing context, and unsupported attributes.
Explore how actors interpret the question, disagree, raise objections, and influence later responses.
Turn decision-critical claims into interviews, surveys, usability sessions, or live tests with appropriate participants.
Useful output makes the next decision or validation step more specific.
These boundaries apply even when the output looks detailed or consistent.
Frequently asked questions
Synthetic users are AI-generated actors conditioned on source evidence and explicit assumptions. They can rehearse research questions, but they are not recruited participants or measured customers.
A persona is usually a static description, while a synthetic respondent generates answers. MiroFish synthetic users are evidence-grounded actors that can react across multiple roles and rounds in a bounded scenario.
No. They can help improve an interview guide and widen the hypothesis set, but only real participants can provide lived experience, natural language, and direct human evidence.
Upload current interviews, reviews, support tickets, sales and win-loss notes, product briefs, positioning, competitor evidence, audience definitions, and the exact decision the research must inform.
No. MiroFish models actor reactions to supplied evidence and scenarios; it does not operate a browser, click through an interface, or replace observed usability testing.
Validate whenever a claim could materially change a product or GTM decision, and whenever the question depends on actual behavior, prevalence, accessibility, lived experience, purchase, or willingness to pay.
Bring the current evidence, a bounded question, and the assumptions your team is willing to challenge.
Related methods and scenarios
Turn market evidence into inspectable research scenarios and next-step questions.
Model how audience groups influence one another across a GTM scenario.
Choose generated rehearsal or direct human evidence for the decision.
