Synthetic respondents are AI-generated representations that produce simulated survey answers or behaviors for a defined participant profile or population. Researchers may create them with language models, demographic attributes, prior survey data, or contextual information. They can explore response patterns, test questionnaires, or supplement early research before collecting answers from people.
For example, a research team might ask synthetic respondents with different profiles to answer the same product survey and compare the distributions. These outputs can reveal assumptions worth testing, but they do not become representative evidence simply because many responses were generated. Results may reproduce biases in the model or source data, and small prompt changes can affect the answers. Synthetic respondents are different from verified human participants and should not be presented as direct substitutes without validation. Teams need to document how the profiles were created, what data grounded them, which model and prompt were used, and how the simulated responses were compared with real observations.