Answers are grounded in this paper and cite their sources.
Research on synthetic respondents has concentrated on fidelity: how closely AI-generated answers match those of human panels. This note asks how organisations use a synthetic panel once they have accepted it. We analyse the usage logs of one provider: 895 survey projects run by 24 enterprise teams between May 2024 and October 2026. Projects fall into two clusters of similar size: testing (52%), dominated by concept screens set up mostly by research functions, and developing (47%), descriptive questions about customers and markets set up almost entirely outside research. Across the data corpus, 70% of projects were set up outside the research function and the median project reached results in 3.9 hours. On this basis, we propose the concept of synthetic micro-research: fast market research on synthetic panels set up primarily by non-researchers, which we hypothesise serves low-stakes decisions that would otherwise have gone untested. The note closes with a research agenda for synthetic research in use and implications for corporate research functions and agencies.
No takes yet. Share an insight, caveat, or question.
Zahradníčková et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: