peer reviewed · Frontiers in Psychology (2025) · Sep 18, 2025
The uncanny valley effect in embodied conversational agents: a critical systematic review of attractiveness, anthropomorphism, and uncanniness
Ștefania Cihodaru-Ștefanache, Ioana R. Podina
In plain English
From 21,897 screened records, 29 studies met the bar for this review of what actually makes an embodied conversational agent feel uncanny. Design features that moved the needle included communication style (humour, empathy, positive phrasing), facial expression, non-verbal cues, voice, customisation and realism, while user gender, age, openness to experience and prior familiarity with agents all shaped the response — but results were mixed and some features increased uncanniness. The authors rated 17 of the 29 studies methodologically weak and 12 moderate, citing unclear randomisation and unvalidated measures, so their practical output is a design checklist rather than a set of rules. Useful as a structured list of things to test on your own users, not as settled guidance.
Related research
- Effects of AI-Powered Embodied Avatars on Communication Quality and Social Connection in Asynchronous Virtual Meetings
- Perception and Monitoring of Sign Language Acquisition for Avatar Technologies: A Rapid Focused Review (2020-2025)
- Would You Rely on an Eerie Agent? A Systematic Review of the Impact of the Uncanny Valley Effect on Trust in Human-Agent Interaction