What an independent study of AI interviewers taught us

Nielsen Norman Group tested two AI interviewers with 10 people and published what went wrong: interviews that ran 13 to 56 minutes, an AI that couldn't tell when a question was done, and constant thanking that one participant compared to a sociopath.

Inqueria already addresses three of these. It sets and states an interview length up front (and this week we widened it after a real participant ran long). It moves on when answers stop adding anything new. And it never flatters.

The fourth, letting participants respond to a summary before moving on, we haven't built yet. It's next on the list.