Whose hotel does the AI recommend? An algorithm audit of reputation signals in LLM-assisted hotel selection

This study presents an algorithm audit of LLM-assisted hotel recommendations, analyzing how various reputation signals influence selection across different models and prompts. Key findings indicate that guest ratings and price significantly affect recommendations, with a top rating increasing selection likelihood by 31.6 percentage points and high prices decreasing it by 30.0 percentage points, while eco-certification is over-weighted. This research emphasizes the need for transparency in LLM decision-making processes, highlighting implications for optimizing generative models and ensuring accountability in AI-driven recommendations.