Signal Invisible to the Machine: Census-Based Audit of AI Restaurant, Cafe and Bar Recommendations in Bali

Summary

This study presents what it calls the first census-based audit of AI venue recommendation systems for food and drink businesses. The authors carried out a full enumeration of 4,776 cafes, restaurants and bars across two areas of Bali, Indonesia, Canggu and Ubud, and evaluated 2,208 search-grounded responses from four commercial AI systems including ChatGPT. The study is motivated by the observation that AI assistants are becoming a primary interface for local discovery even as little is understood about which venues they surface. This is framed as a problem with direct revenue consequences for small businesses. The researchers suggest this auditing approach could extend to other regions or industries.