Talking surveys: How photorealistic embodied conversational agents shape response quality, engagement, and satisfaction

In plain English

Eighty UK participants completed two psychometric questionnaires either with a photorealistic video-avatar interviewer or with a text chatbot, producing 2,265 responses. The avatar condition produced significantly more informative and detailed answers and higher, more time-efficient engagement — but satisfaction did not improve, with turn-taking delays and uncanny valley reactions named as the drag. A useful correction to the assumption that embodiment improves everything at once: it can improve the data you collect while leaving how people feel about the encounter unchanged.