Whose hotel does the AI recommend? An algorithm audit of reputation signals in LLM-assisted hotel selection
A new study published on arXiv audits the recommendation signals used by large language models (LLMs) in hotel selection. The research found that guest ratings and price are the most influential factors, similar to human preferences, while other signals like eco-certification are over-weighted and management responses are ignored. Notably, the position of a hotel on a list, a content-free artifact, significantly impacts recommendations, suggesting a need for optimization and accountability in AI infomediaries. AI
IMPACT Highlights the need for transparency and optimization in LLM-driven recommendation systems, particularly in consumer-facing applications like travel.