# AI Propaganda factories with language models

## TL;DR

This paper demonstrates that small language models can generate coherent political propaganda, revealing new challenges for detection and defense strategies against automated influence campaigns.

## Contribution

It shows that small language models can produce effective propaganda and that detection should focus on conversation patterns rather than model access.

## Key findings

- Persona design influences behavior more than model size
- Engagement increases ideological adherence and extremism
- Automated influence campaigns are feasible with small models

## Abstract

AI-powered influence operations can now be executed end-to-end on commodity hardware. We show that small language models produce coherent, persona-driven political messaging and can be evaluated automatically without human raters. Two behavioural findings emerge. First, persona-over-model: persona design explains behaviour more than model identity. Second, engagement as a stressor: when replies must counter-arguments, ideological adherence strengthens and the prevalence of extreme content increases. We demonstrate that fully automated influence-content production is within reach of both large and small actors. Consequently, defence should shift from restricting model access towards conversation-centric detection and disruption of campaigns and coordination infrastructure. Paradoxically, the very consistency that enables these operations also provides a detection signature.