What GEO means for a hotel
ChatGPT, Claude and Gemini learn about a hotel three ways: what their crawlers collect from your site, structured data that states who and where you are, and what other sites (Google Maps, TripAdvisor, Pantip) say about you. GEO makes all three complete and consistent: a Hotel schema with name, address, phone and coordinates; a robots.txt that does not disallow GPTBot or ClaudeBot; an llms.txt that sums up the hotel in 20 lines; and facts as text, not images.
Why it matters
- Measured on a Bangkok hotel, all three AIs got the walking distance to the main venue wrong by 3x, because the site never stated the number, so the AI guessed
- GEO is 30% of the overall score, equal to SEO, because in 2026 many guests ask an AI before opening Google
- Quite a few hotels block AI crawlers in robots.txt without meaning to (a template that disallows every bot), then wonder why ChatGPT never mentions them
What the free scan checks for GEO
This area's score is the share of weight passed (warnings count half). Heavier items matter most to Google and AI.
| # | Check | Weight | Hotels passing |
|---|---|---|---|
| 1 | Hotel schema with name, address, phone Fix: Add JSON-LD @type Hotel: name, address, telephone, geo, checkinTime, amenityFeature, priceRange, sameAs | 16 | - |
| 2 | Name, address, phone readable on the page Fix: Show full name, address and phone as text (not an image) in the footer of every page | 12 | 29% |
| 3 | Location / province stated clearly Fix: Put the province or area in the title, H1, first paragraph and schema address | 12 | 86% |
| 4 | Identity links (sameAs / social) Fix: Add sameAs to Facebook, Instagram, Google Maps, TripAdvisor so AI can match them to the same hotel | 8 | - |
| 5 | Review rating in schema Fix: Add aggregateRating from real reviews (Google/Booking) in the Hotel schema | 8 | - |
| 6 | AI crawlers not blocked Fix: To be known by ChatGPT/Claude/Gemini, never Disallow GPTBot, ClaudeBot or Google-Extended | 12 | 100% |
| 7 | llms.txt Fix: Create /llms.txt: a 20-line summary of the hotel, location, highlights, rooms, rates and key links | 6 | 57% |
| 8 | Facts AI uses to answer Fix: State facts as plain text: check-in 14:00, check-out 12:00, breakfast, shuttle, distance to the airport and landmarks | 10 | 57% |
| 9 | About page Fix: Add a page with history, owners, opening year, room count and awards; AI uses it as a reference | 6 | 43% |
| 10 | og:site_name (brand name) Fix: Set og:site_name to the hotel name spelled the same everywhere | 4 | 57% |
What the full system adds
The full system asks all three AIs every 2 weeks with real guest questions, measures how often you are mentioned, at what position, who beats you, and whether the AI has your facts right (rates, distances, facilities), then creates tasks for the team to fix the site.
FAQ
Is it risky to let AI crawlers in?
For a hotel site with only public information, no. The upside is that AI knows you; the downside is close to none, unlike sites with copyrighted content.
What is llms.txt?
A plain text file at /llms.txt summarising who the site is, what it offers and the key links. 20 lines is enough; several AIs read it first.
Why does AI recommend a competitor further away?
Because the competitor gives the AI more readable data: schema, many reviews, Pantip mentions. AI picks by confidence in the data, not by real distance.