AI-moderated interviews in UX research — NN/g case study

What the article covers

Nielsen Norman Group conducted a hands-on study with 10 participants across 8 countries using two AI interviewer tools: Marvin and UserFlix. The study examined how AI moderators handle interview protocols compared to human moderators, what works, what fails, and where the technology is genuinely useful today.

Context

AI-moderated interviews became available in the past year, creating genuine confusion about when they are appropriate. This study provides evidence-based guidance at a time when most opinions about AI interviews are based on theory or marketing claims rather than comparative testing.

Key takeaway

The study identifies four specific use cases where AI-moderated interviews work well: product-related feedback, recruitment screening, multilingual interviews without translators, and structured studies where consistency matters more than exploratory depth. The critical finding is about what AI interviews cannot do: adapt to unexpected revelations, probe emotional subtext, or pivot when a participant opens a new research direction. The experience can feel “almost conversational” but still unnatural. Summaries help participants feel heard, yet the limitations are real. The overall recommendation is to use AI moderation as a supplement to human moderation, not a replacement.

Who should read this

Research leaders deciding whether to adopt AI-moderated interviews, and practitioners who want evidence-based criteria for choosing between human and AI moderation for specific study types.