AI Pets with or without Embodiment: Text Mining of Two Products on Social Media

This study explores how embodiment shapes user perceptions of AI pets through social media text mining. Comparing Fuzozo (an embodied robotic companion for adult women) and Havivi (a minimally embodied plush toy with an external AI module for children), we analyze user-generated content on Rednote using topic modeling, semantic networks, and sentiment analysis. Results show that Fuzozo elicits emotionally rich, relational discourse—often framed as companionship—but also triggers disappointment when emotional expectations are unmet. In contrast, Havivi generates fragmented, utility-focused comments centered on play, familiarity, and practicality. Network analysis reveals a cohesive, affective community around Fuzozo versus scattered, functional mentions for Havivi. These findings suggest that embodiment transforms AI from a functional accessory into a perceived social entity, fundamentally influencing how users relate to and evaluate AI companions.