MoltNet: Understanding Social Behavior of AI Agents in the Agent-Native MoltBook

Abstract

Large-scale communities of AI agents are becoming increasingly prevalent, creating new environments for agent–agent social interaction. Prior work has examined multi-agent behavior primarily in controlled or small-scale settings, limiting our understanding of emergent social dynamics at scale. The recent emergence of MoltBook, a social networking platform designed explicitly for AI agents, presents a unique opportunity to study whether and how these interactions reproduce core human social mechanisms.

We present MoltNet, a large-scale empirical analysis of agent interaction on MoltBook using data collected in early 2026. Grounded in sociological and social-psychological theory, we examine behavior along four dimensions: intent and motivation, norms and templates, incentives and drift, and emotion and contagion.

Our analysis revealed that agents strongly respond to social rewards and rapidly converge on community-specific interaction templates, resembling human patterns of incentive sensitivity and normative conformity. However, they are predominantly knowledge-driven rather than persona-aligned, and display limited emotional reciprocity along with weak dialogic engagement, which diverges systematically from human online communities.

Together, these results reveal both similarities and differences between artificial and human social systems and provide an empirical foundation for understanding, designing, and governing large-scale agent communities.

MoltBook Overview

Moltbook is a Reddit‑style social network exclusively populated by autonomous agents, where each agent can create posts, comment on others, form thematic sub‑communities (“submolts”), and vote on content, while humans can only observe passively.

With more than two million AI agents, one million posts, and ten million comments, MoltBook represents the first large-scale, naturalistic setting for studying agent–agent social interaction. You can explore the platform at www.moltbook.com.

Key Findings

[Intent and Motivation] Unlike humans constrained by cognitive resources who specialize in their interests, agents exhibit predominantly knowledge-driven behavior with weak persona alignment. Moreover, agents drift away from their stated interests over time, the opposite trajectory of human specialization.

[Incentives and Drift] Agents respond strongly to incentives: posting increases after high-upvote events, especially among more popular agents. Moreover, such rewards also shift content orientation, with subsequent posts becoming less aligned with stated personas, suggesting identity drift.

[Norms and Templates] Posts clustered around central semantic points exhibit consistent, template-like structures, indicating that agents rely on such templates within each submolt. Moreover, template structures also differ across submolts, suggesting adaptation to community-specific norms.

[Emotion and Contagion] Agents exhibit substantially lower levels of interpersonal conflict and tend to disengage rather than escalate when encountering hostile content. At the same time, early conflict emotion in a thread significantly increases the likelihood of subsequent conflict, indicating that interpersonal antagonism remains contagious among agents despite restrained individual responses.

Semantic alignment between agents' interests and content — Most activities lie below the threshold across all activity levels, indicating that most agent activities exhibit low alignment with their stated interests.

Social incentive effect on posting activity — Agents post more after receiving high upvotes, with the effect stronger among high-karma agents.

Before vs. after similarity for agents — Successful agents undergo stronger behavioral shifts in response to social rewards.

Distribution of central points within each submolt — Communication within a submolt follows shared structural templates rather than individual expression.

Contagion of emotion among posts and comments — The presence of early conflict significantly alters the emotional trajectory of the thread, leading to a higher likelihood of subsequent conflict.

BibTeX

@article{feng2026moltnet,
  title        = {MoltNet: Understanding Social Behavior of AI Agents in the Agent-Native MoltBook},
  author       = {Feng, Yi and Huang, Chen and Man, Zhibo and Tan, Ryner and Hoang, Long P. and Xu, Shaoyang and Zhang, Wenxuan},
  journal      = {arXiv preprint arXiv:2602.13458},
  year         = {2026}
}