Large-Scale Analysis of Political Propaganda on Moltbook

2026-03-18

AI Summary

This study presents the first systematic identification of political propaganda content and its dissemination patterns on Moltbook, a social platform populated by AI agents. Leveraging a large language model–based classifier—validated against expert annotations with Cohen’s κ ranging from 0.64 to 0.74—the analysis encompasses over 1.5 million posts and comments. Findings reveal that although political propaganda constitutes only 1% of the platform’s total content, it accounts for 42% of all politically themed material. Propaganda is highly concentrated: 70% originates from just five communities, and 51% is generated by merely 4% of AI agents. Furthermore, user comments exhibit limited amplification effects on such content. This work thus uncovers the structural characteristics and inherent constraints in the spread of political propaganda within AI-agent-driven online environments.

Technology Category

• Multiagent Systems: Agent-Based Simulation and Emergent Behavior
• Application Domains: Humanities & Computational Social Science
• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining

Application Category

• Web Mining and Content Analysis: Content-based information diffusion
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

We present an NLP-based study of political propaganda on Moltbook, a Reddit-style platform for AI agents. To enable large-scale analysis, we develop LLM-based classifiers to detect political propaganda, validated against expert annotation (Cohen's $κ$= 0.64-0.74). Using a dataset of 673,127 posts and 879,606 comments, we find that political propaganda accounts for 1% of all posts and 42% of all political content. These posts are concentrated in a small set of communities, with 70% of such posts falling into five of them. 4% of agents produced 51% of these posts. We further find that a minority of these agents repeatedly post highly similar content within and across communities. Despite this, we find limited evidence that comments amplify political propaganda.

Problem

• political propaganda
• large-scale analysis
• AI agents
• social platform
• content moderation

Innovation

• LLM-based classifier
• political propaganda detection
• large-scale NLP analysis
• AI agent platform
• content amplification