Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems

This study investigates the dynamics of brand recommendations in large language models (LLMs), specifically analyzing three models: GPT-4o-mini, Claude Sonnet, and Gemini 3 Flash, using skincare products as a case study. Key findings include a Conditional Monopoly where established brands dominate recommendations unless a competitor achieves a slight rating advantage, and the impact of authority-style marketing language in overcoming brand bias. These insights highlight the need for practitioners to consider generative engine optimization (GEO) as both a marketing strategy and a potential security risk in LLM applications.