Artificial Voices: A Comparative Analysis of Brand Identity Preservation in LLM-Generated Content

Brand communication is a strategic asset for building identity and trust, yet the evolution of Large Language Models (LLMs) introduces a radical shift in how these narratives are constructed. This study investigates the ability of LLMs to replicate the distinctive communicative styles of global brands, specifically Nike and Adidas. The primary objective is to evaluate the similarity between authentic and generated texts, determining the extent to which algorithmic mediation can preserve a brand’s rhetorical traits. The research employs a two-step methodology using a dataset derived from six digital channels (including social media, newsrooms, and e-commerce). First, the model was trained on contextual data to generate content for new products; second, the training was refined using automatically extracted quantitative linguistic constraints. The findings offer a critical perspective on the relationship between human and artificial language, highlighting the potential and limitations of LLMs in maintaining brand identity through linguistic automation.