An AI-generated person who answers research questions in place of, or before, a real respondent. It is the survey and panel industry's word for a synthetic user. Synthetic individual, simulated people and AI persona mean much the same thing. A digital twin is different, because it copies one specific real person.
Respondent is the survey and market research word for the person who answers. A synthetic respondent fills that seat with an AI-generated person. You will see the term most from panel companies and survey platforms, while product and UX teams more often say synthetic user. The idea is the same: a simulated member of your audience who answers your questions.
The wording matters less than what sits behind each respondent. Whether each one is built from evidence you can trace, and whether they stay the same person for the whole study, decides how far you can trust what they say.
Platforms that offer synthetic respondents split into two groups.
Survey-style. The platform generates hundreds or thousands of respondents, asks each a fixed questionnaire, and reports percentages: 62% prefer option B. This suits questions with a small set of possible answers. It hides the reasoning, and it hides how much the numbers depend on the model’s assumptions.
Interview-style. The platform builds a smaller number of detailed individuals and interviews each one in depth, following up on what they say. The output is the reasoning: why a group rejects a concept, which objection comes up first, where two groups disagree. Candor works this way. The essay synthetic interviews are not synthetic polls sets out the difference in full.
Four questions separate tools you can build on from tools that write fluent guesses.
What is each respondent built from? A description in a prompt, or evidence about the audience you can inspect.
Does each respondent stay the same person? A respondent who forgets what they said two questions ago is not a person, just a model answering again.
Do respondents disagree? If every respondent likes every concept, the tool is measuring the model’s politeness.
Can you see what was assumed? A good platform shows which parts of a respondent, and which findings, rest on evidence and which it filled in.
The guide to synthetic research tools applies a longer version of this test to 24 platforms, and Traditional vs synthetic user research compares synthetic respondents with recruited ones.
For the method behind synthetic respondents, read what synthetic user research is. To see where synthetic work fits beside real fieldwork, read where synthetic research fits in your workflow.
A real panel is made of recruited people who are paid to answer, which takes time to field and costs money per complete. Synthetic respondents are generated, so a study can run the same day and be repeated as often as you like. The trade is certainty. A real respondent's answer is a fact about that person. A synthetic respondent's answer is an estimate of how a type of person would reason, only as good as the evidence behind it.
Not for decisions that need a measured number, such as market share or a precise purchase intent score. Synthetic respondents are better at the questions that come before a survey: which problems matter, why people react the way they do, which of ten options deserve a real test. Many teams use them to decide what to put in front of a real panel, not to skip the panel.
Ask four things. What is each respondent built from, and can you see it? Does each respondent stay the same person across a study, remembering what they said? Do respondents push back and disagree, or agree with everything? And can the platform show which findings rest on evidence and which it assumed? Tools that cannot answer these tend to produce fluent answers that are hard to trust.
Both. Some platforms generate large numbers of respondents to answer survey questions and report the percentages. Others, including Candor, run in-depth interviews with a smaller number of individuals and report on the reasoning behind their answers. Interviews show why people think what they think, which percentages alone cannot.
Start with a free project, or walk through Candor with us first.