arXiv AI recent: Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
Researchers studied brand dynamics in large language model (LLM) recommendations using skincare products.,They found that well-known brands get recommended 100% of the time when all produ...
The study used three commercial LLMs (GPT-4o-mini, Claude Sonnet, Gemini 3 Flash) and a robustness check on search goods.,The experiments found that a Conditional Monopoly exists where well-known brands dominate recommendations, but this dominance disappears with a small rating advantage for a co...