Digital-twin accuracy depends on both how accuracy is measured and what question the twin is asked. Using multiple survey datasets, the research shows why strong aggregate results can hide weak individual-level accuracy. It then introduces an R_-based answerability screen for evaluating new questions when human answers are unavailable. The screen helps researchers identify which synthetic results are more defensible and which questions still require human research.