LLM Silicon Sampling Fails to Cover Human Respondent Profiles
October 23, 2026
The Respondent Coverage Gap (RCG) metric reveals that LLM-generated synthetic populations fail to represent the full spectrum of human answer profiles. Even when models match individual answer distributions, they leave human respondents uncovered by producing combinations of answers not found in real datasets.
HOW THIS AFFECTS YOU
●
builderBe cautious when using LLM-generated populations for market research or user profiling.
●
researcherYou should use RCG instead of simple distribution matching to evaluate synthetic data quality.