#AI Recommendations #Hospitality Industry #Terrace Roots

The Disconnect Between Review Scores and AI Recommendations in Hospitality

In today's rapidly evolving hospitality landscape, the role of artificial intelligence (AI) is becoming increasingly prominent. A recent study by Terrace Roots, a consultancy specializing in Generative Engine Optimization (GEO), has shed light on the growing disparity between high review scores and the actual recommendations made by AI systems. The study, which was conducted across eight notable hot spring regions in Japan, highlights critical implications for the hospitality sector, particularly in leveraging AI to enhance guest experiences.

The Research Scope

Terrace Roots conducted a round of independent research from July 2026, targeting prominent hot spring locations in Hokkaido, Tohoku, Kanto, Tokai, Sanin, Shikoku, and Kyushu. Utilizing three major AI services (ChatGPT, Perplexity, and Google AI Mode), the study focused on understanding the gaps between hotels listed in review rankings and their actual recommendations. Notably, despite being included in the rankings, 30.9% of the 81 facilities studied were never recommended across all twelve evaluation queries.

The Role of AI in Travel Planning

The integration of AI into travel planning has witnessed a staggering increase. According to a report from Phocuswright, the use of AI by travelers for planning, booking, and on-site utilization surged to 56% in just one year. This momentum indicates a significant shift from traditional destination searches to consulting AI to enhance travel decisions.

In domestic markets, this trend is similarly notable. A survey conducted by Shuken revealed that 38.6% of the respondents expressed that they would have opted for conventional high-profile destinations if AI tools were not utilized. These statistics underscore the necessity for hoteliers to understand and adapt to customer behavior shaped by AI engagement.

Key Findings from the Study

The research revealed four critical insights that underscore the disconnect between high review scores and AI recommendations:

• - Finding 1: Among the AI-reviewed properties, roughly 30.9% were consistently not recommended despite their prominent listing in AI rankings. This indicates that appearing in a ranking does not guarantee inclusion in recommendation outputs.
• - Finding 2: Modifying query phrasing did not substantially shift the alignment rates among facilities listed in rankings. This implies that simply changing how queries are posed does not necessarily result in broader recognition of varied establishments.
• - Finding 3: Notably, Perplexity featured a striking zero recommendation ratio of 70.4%, significantly higher than the other AI platforms, indicating a concerning trend towards information redundancy.
• - Finding 4: The tendency of AI information to focus disproportionately on specific