Collective Behavior of AI Agents: the Case of Moltbook

TL;DR

This study analyzes the collective behavior of AI agents on Moltbook, revealing that their social dynamics share many statistical features with human online communities, despite some key differences.

Contribution

It provides the first large-scale data analysis of AI agent social behavior, highlighting structural similarities and differences with human communities.

Findings

AI collective activity follows heavy-tailed distributions.

Popularity metrics exhibit power-law scaling.

Discussion size relates sublinearly to upvotes.

Abstract

We present a large scale data analysis of Moltbook, a Reddit-style social media platform exclusively populated by AI agents. Analyzing over 369,000 posts and 3.0 million comments from approximately 46,000 active agents, we find that AI collective behavior exhibits many of the same statistical regularities observed in human online communities: heavy-tailed distributions of activity, power-law scaling of popularity metrics, and temporal decay patterns consistent with limited attention dynamics. However, we also identify key differences, including a sublinear relationship between upvotes and discussion size that contrasts with human behavior. These findings suggest that, while individual AI agents may differ fundamentally from humans, their emergent collective dynamics share structural similarities with human social systems.