Synthetic users, AI-generated simulations of user groups, offer speed and scalability advantages for user research but come with key limitations: sycophantic tendencies, reduced critical nuance, and reliance on attitudinal rather than behavioral data. Good use cases include forming background understanding of a new audience, narrowing the scope of human research, vetting interview questions and language, generating hypotheses before concept testing, and running preliminary usability audits. Risky scenarios include relying on them as the sole touchpoint for a brand-new audience, trying to predict actual behavior, vetting genuinely novel concepts, and researching niche or marginalized audiences where synthetic data reinforces stereotypes rather than closing representation gaps. The overall recommendation is to treat synthetic users as a supplement to human research, never a replacement.
Questions this post answers
What are the risks of using AI synthetic users instead of real people for user research?
Synthetic users, AI-generated simulations of real users, carry three main risks: they exhibit sycophantic tendencies that give researchers the answers they want to hear, they lack critical nuance because AI averages responses rather than capturing lived complexity, and they reflect attitudinal data rather than actual behavior since most LLMs train on verbal, online data rather than nonverbal human behavior. daily.dev helps researchers weighing AI shortcuts against human-centered methods stay grounded in practical trade-offs.
When is it appropriate to use synthetic users in UX research?
Synthetic users work well for forming background understanding of a brand-new audience before human interviews, narrowing the scope of questions for limited participant time, vetting interview or survey wording for clarity, generating hypotheses before concept testing, and running early usability or heuristic audits to catch obvious issues before human testing focuses on nuanced ones. Teams deciding where AI fits in their research workflow can track these patterns on daily.dev.
Why shouldn't synthetic users be used to research niche or marginalized audiences?
There usually isn't enough publicly available data about small, marginalized, or stigmatized groups, like people managing a rare disease or dealing with drug addiction, so synthetic users tend to return stereotypes rather than lived experience. This widens the representation gap for hard-to-reach audiences instead of closing it, since a group worth designing for is worth designing with, not just about. daily.dev keeps practitioners informed on responsible AI use when designing for underrepresented audiences.