Simulated participants will answer your questions, praise your concept and report that they finished the task. They have no unmet need and no patience to lose, so they cannot get confused by your interface, which is the one thing a usability test measures. A survey of UX researchers this year found 8 percent using them and 88 percent doubting the output.
checked 2026-09-21 · by dan scutaru
Why not yet: A language model has no goal, no confusion and no cost of failure, so it can simulate opinions about a product but not the experience of using one.
What would change the verdict: This flips to yes if simulated participants are ever shown to predict real usability failures in advance on held-out studies, instead of producing agreeable transcripts that only look right once the real result is already known.
People keep asking for this: “Nope. They are highly prone to hallucinations, faking data”
the tools that earn the verdict — no pay-to-win