When AI answers hotel questions, which websites does it cite?
A Vektos study of 14 Cyprus hotels and 2,258 source references in OpenAI and Google Gemini answers.
- 3.4%
Hotel websites were 3.4% of the sources AI cited. Booking platforms were 44.9%.
Of 2,258 recorded source references · March 2026 - About 13×
Booking platforms were cited about 13 times more often than hotel websites.
44.9% ÷ 3.4% · same references - 0%–77.5%
Some hotels appeared in most AI answers; others in none.
Measured mention rate per response, across the 14 hotels · April 2026
Next edition: December 2026. A wider check of Cyprus hotels, published before the Cyprus Hotel Summit.
Hotel websites were 3.4% of recorded sources.
Share of 2,258 recorded source references in the measured answers · March 2026.
A wide range within the same 14 hotels.
Metric: measured mention rate per response.
Not a recommendation rate or a current Discovery score, and not measured with today’s Discovery method.
What this means for your hotel.
When guests ask AI where to stay, AI often cites booking platforms and other websites, and less often the hotel’s own site. That doesn’t prove a lost booking, but it means your direct channel may be missing from the answers guests read.
3 things to check first.
- Ask ChatGPT and Gemini the questions your guests ask. Is your hotel there, and which website do they link?
- Check that your Google profile shows your own website and your own price.
- Check that your hotel’s facts (name, stars, rooms, season) match everywhere: website, Google, booking platforms.
What this measures.
This study counts displayed source references in measured AI hotel answers. The April analysis also recorded whether each hotel was mentioned in each measured AI response; that historical mention-rate metric is distinct from recommendation position.
It does not measure clicks, bookings, traffic, revenue or whether a source caused the AI answer.
Limitations.
- Historical, point-in-time evidence. AI outputs can vary over time. Publication month is not an exact measurement date.
- A limited sample of 14 hotels, primarily in Ayia Napa and Limassol—not every Cyprus hotel.
- Standardized English travel-intent questions; findings do not establish results for other languages or all traveller questions.
- Recorded source references are not clicks, bookings, traffic or revenue. They do not establish user behaviour or causation.
- The mention-rate range uses an earlier mention-based method, not the current Discovery recommendation methodology.
See where your hotel stands.
We’ll email you 3 findings about your hotel within 2 working days, free. Or get the full six-page Discovery report.