My Favorite Streamer is an LLM: Discovering, Bonding, and Co-Creating in AI VTuber Fandom

Paper Title

My Favorite Streamer is an LLM: Discovering, Bonding, and Co-Creating in AI VTuber Fandom

Publication Info

• Topic area: Human–AI interaction in entertainment and fandom.
• Keywords: AI VTuber, Neuro-sama, parasocial relationships, co-creation, fan economy, participatory culture, monetization, authenticity, human–AI intimacy.

Background and Problem

• Problem / challenge: Limited understanding of audience engagement with fully autonomous AI VTubers, including parasocial dynamics, community formation, and monetization.
• Significance: AI VTubers challenge traditional notions of authenticity and participatory culture, offering insights into human–AI relationships and platform economies.
• Motivation and related work: Prior research has focused on human VTubers and AI as supportive tools, leaving unexplored the dynamics of fully autonomous AI entertainers. This study addresses gaps in understanding audience attraction, attachment, and financial support mechanisms.

Solution

• Proposed approach: A qualitative study of Neuro-sama, combining surveys, interviews, and livestream interaction log analysis to explore audience engagement.
• Novelty:
  1. First systematic study of fan communities centered on AI VTubers.
  2. Introduction of concepts like "consistency-as-authenticity" and "real-time co-performance commodification."
  3. Design implications for AI-mediated entertainment balancing fairness, persona stability, and ethical safeguards.
• Procedure and key techniques:
  • Survey: 334 fans to capture motivations, parasocial relationships, and financial behaviors.
  • Interviews: 12 dedicated fans to explore emotional narratives and attachment.
  • Interaction log analysis: Comparative study of chat and SuperChat data from Neuro-sama and human VTubers.

Results

• Concrete findings:
  • 96% of fans discovered Neuro-sama via algorithmic recommendations; unpredictability and community interaction were key attractions (92% and 90% rated as important).
  • Parasocial relationships are built on emotional connection (99% fondness), cognitive engagement (76% observing behavioral patterns), and behavioral interaction (69% active participation).
  • SuperChats for Neuro-sama are predominantly proactive (85%), enabling fans to co-create content, unlike human VTubers where reactive messages dominate.
  • Neuro-sama’s monetization model shows higher paid conversion rates (1.59%) and income stability (Gini coefficient: 0.24) compared to human VTubers.
• Advantage over baselines: Neuro-sama’s fans engage more actively in shaping content, with higher proactive chat and SuperChat proportions compared to human VTubers.
• Experiments / evaluation:
  • Surveys measured motivations and parasocial scales.
  • Interviews provided qualitative depth on emotional and cognitive attachment.
  • Interaction logs analyzed chat and SuperChat dynamics across AI and human VTubers.
• Limitations and future work: Focused on a single AI VTuber (Neuro-sama), limiting generalizability. Future research should include cross-cultural comparisons, broader AI VTuber samples, and qualitative data from human VTuber fans.

Summary

This study explores how audiences engage with AI VTubers like Neuro-sama, revealing novel dynamics in attraction, parasocial relationships, and monetization. Fans are drawn by unpredictability and community interaction, develop emotional bonds through anthropomorphic projection, and sustain engagement via co-creation opportunities enabled by SuperChats. Neuro-sama’s consistent persona redefines authenticity, while her monetization model demonstrates higher stability and efficiency compared to human VTubers. These findings extend theories of participatory culture and mediated authenticity, offering design implications for balancing fairness, persona stability, and ethical safeguards in future AI-mediated entertainment.