How Political narratives are shaped and distributed by LLMS (Russian-Origin narratives)

The study introduces a focus on how large language models distribute and shape contested political narratives across conflicts. It advances the field by reframing LLMs as generative visibility regimes (extending Foucault's regime of truth and Brighenti's regime of visibility), systems that author what becomes visible, thinkable and invisible (Gillespie, 2024). By comparing outputs from five large language models across English and Norwegian and three conflicts, the research introduces a cross-model, cross-lingual approach to studying narrative reproduction, offering new conceptual tools for understanding AI-mediated geopolitics and informing policies that safeguard democratic information environments. The project is structured as an article-based dissertation with three interconnected studies and a synthesizing kappa.