What we tested
An overview of the research design, including how we structured parallel studies with synthetic and real respondents.

How responses differed
Key findings on where synthetic participants aligned with, or diverged from, human data and what our results suggest about the strengths and limitations of using AI-generated respondents in various research scenarios.

Workflow considerations
A look at how survey logic, branching, and object detection influenced participant experience and outcomes.

Practical takeaway
Where synthetic inputs can realistically support qualitative and quantitative goals, and where caution is still warranted.