An LLM-driven framework for strategic writing style transformation in cyber influence operations

Abstract

In the evolving landscape of cyber and cognitive warfare, language has emerged as a decisive instrument for shaping perception and influencing digital audiences. Effective communication on social media requires not only timely information delivery but also stylistic adaptability to maximize message reach and resonance. This paper introduces a Large Language Model (LLM)-based framework designed to optimize writing style transformation for strategic influence operations in online environments. Our system converts original textual content across three key styles - Humorous, Analytical, and Critical - spanning five thematic domains: Culture, Sports, Entertainment, Technology, and Politics. Through controlled style modulation, this method aims to enhance both information diffusion and positive engagement (“active dissemination”) while preserving message intent and factual coherence. We propose a multi-stage pipeline integrating stylistic control, semantic alignment, and evaluative feedback to select the optimal style for each context. Empirical evaluations, including pairwise statistical tests and diffusion analysis, demonstrate that style transformation significantly impacts audience interaction patterns and sentiment trajectories. The results can serve as a foundational tool for cyber influence strategists, enabling adaptive, ethically guided, and high-impact communication in the dynamic information battlespace.