SYNTHETIC USERS VS. HUMAN RESPONDENTS: SAME CONCLUSION, LESS VARIATION

Summary from the TLDR Design newsletter

A comparison of 120 synthetic respondents with 604 humans produced the same broad conclusion about terminology for UX work, but AI-generated responses showed dramatically less variation. Synthetic participants converged on roughly a dozen labels while humans produced more than 50, demonstrating a form of mode collapse that can erase unusual perspectives and differences between groups. Synthetic users may help generate hypotheses or identify broad patterns, but they cannot replace human research when unexpected responses, minority perspectives, or natural variation matter.