Article|12 August 2026
4 min read

What is an AI-moderated interview?

Share

An AI-moderated interview is a survey in which an AI leads the conversation: it asks questions, listens and follows up – like a good interviewer, only automated. The term is new; the method behind it is surprisingly simple. Here we explain how it works, how it differs from questionnaires and classic interviews – and how many conversations it takes.

The definition in three sentences

In an AI-moderated interview, an artificial intelligence takes on the role of the interviewer. Participants receive a link and hold the conversation whenever and wherever suits them – on their smartphone or computer, in their own language. The AI asks open questions, responds to every answer and follows up until the real reason has been named.

The method thus combines what used to be mutually exclusive: the depth of an individual interview and the breadth of a survey. A hundred in-depth interviews, conducted in days instead of months.

How does an AI-moderated interview work?

It starts with your question. From it, we derive the guiding questions for the conversation – the AI does the rest. An example from an interview on customer churn:

AI: “You said you switched because of the price. Was the price the trigger or the final push?”

Participant: “To be honest, the final push. Something else had been annoying me for a while.”

This is exactly where a questionnaire ends – and where an AI-moderated interview only gets started. On average, each interview has more than 11 rounds of dialogue.

How does it differ from a questionnaire?

A questionnaire cannot follow up. Standard questions lead to standard answers – the why stays hidden. And the first answer is rarely the complete one: people name the most obvious reason first, not the real one. An AI-moderated interview stays with it and still scales as easily as a survey.

And how does it differ from in-depth interviews and focus groups?

Not in the quality of the conversation – in the economics. Classic in-depth interviews and focus groups need scheduling, moderation, transcription and weeks of analysis. That is why most projects stop at twenty conversations. An AI-moderated interview has no such limit: a hundred or more conversations run in parallel, without appointments and without transcription.

The method has also been studied scientifically: in a study by the economists Chopra and Haaland, an AI conducted 395 interviews in conversational format. Over 80% of participants rated the experience positively – more than half preferred the AI-moderated interview to a human conversation. What this means for organisations is shown in our article How AI-moderated interviews help organisations.

How many interviews does it take?

Our analyses of 1,576 conversations showed: 90% of the core themes are first mentioned after 11 to 23 conversations, and after around 30 practically all of them appear. Every further interview strengthens the evidence – if you want to compare subgroups or need particularly robust figures, you interview more people. This is consistent with research on saturation in qualitative studies (Guest et al. 2006, Hennink & Kaiser 2022).

What comes out at the end?

Not a mountain of text, but a navigable report: all answers are organised by theme, and every insight shows how many respondents support it – backed by quotes from the conversations. Subgroups can be compared with a click, and follow-up surveys make change over time visible.

Frequently asked questions

What is the response rate? In a typical survey, 8 in 10 interviews that are started are completed – a conversation that listens feels different from a form.

How quickly are results available? Typically, 80% of responses are in within 24 hours; the analysis runs automatically alongside.

And data protection? All data is hosted 100% in Switzerland, with no third countries – compliant with the GDPR and the Swiss FADP. Confidential and pseudonymised on request.

Would you like to see what this looks like for your question? We'll show you live in 20 minutes.