Research
AI Hotel Research
Data-driven studies on how AI search recommends, ranks, and cites hotels — plus the cross-industry AI-search infrastructure methodology behind them. All research is open and independently published.
Looking for the fun stuff? Visit the Playground6 results
MethodologySeptember 2026
What Selects Google AI Mode's Entity Wrapper
44 Places types × 10 query forms × 4 cities, replicated at +2h
0.876 → 0.198 same noun, list vs why76.2% of citations wrapped (T1)0.086 geographic areas (T2)
MethodologySeptember 2026
AI Traffic Is Undercounted
Four measurements of one hotel, and they disagree by 35x
24 of 35 AI arrivals, no AI signal96.4% Google clicks branded0.26% Discovery-query CTR
MethodologySeptember 2026
AI Mode Stopped Linking Publishers
Decoding the searchviewer svid back to an entity
99.96% Citations wrapped100.000% Decode rate6,792 Entities recovered
MethodologyAugust 2026
Raw-Stream Hidden Signals
What ChatGPT actually sends, and doesn’t
141 ChatGPT citation-competitor groups0/146 ChatGPT brand priors0/146 ChatGPT model escalation
MethodologyAugust 2026
How many apps are there in Claude and ChatGPT?
The first-party census: 1,735 ChatGPT apps, 1,375 Claude connectors, and a travel vertical instrumented end to end
1,735 ChatGPT apps (anon, FR)1,375 Claude connectors5% Registry-store overlap
MethodologyJuly 2026
The Cost of AI Crawling
119,609 AI-bot hits, an OpenAI/Claude duopoly, and a log table that ballooned to 5.6 GB
119,609 AI-bot requests81% OpenAI + Claude5.6 GB One log table