Travellers no longer start with a list of links
A guest can ask:
"We're spending a weekend in Limassol. Where should we stay near the old town and waterfront?"
Replies can combine hotel suggestions, descriptions, maps, current web information and source links.
OpenAI says ChatGPT may automatically search for current information and include citations.1 Google says Gemini’s trip planning draws real-time locations, reviews and prices from Maps, Flights and Hotels.3
01The question decides what is relevant
The same town can prompt different needs:
- Where should a couple stay in Limassol for a weekend?
- What are the best family hotels in Limassol?
- Where can I stay near Limassol Old Town?
- Which beachfront hotels are good value?
- Tell me about Asteri Harbour Hotel.
Location, traveller type, budget and prior hotel awareness change which properties fit. A hotel can appear for one question and not another.
One “best hotels in Limassol” query tells management little.
02The assistant may look things up
AI assistants can go beyond training data. When ChatGPT uses a search partner, OpenAI says it typically rewrites the question into targeted queries.1 Google’s Gemini developer documentation describes deciding whether search helps, running searches, combining results and returning a cited response.2
Think of an answer as assembled at the time of the question, not retrieved from one fixed hotel database. Results vary by product, question, country and feature.
03Your website is only one source among many
An assistant can draw on several sources, so “Is our website on Google?” is too narrow a question.
When AI answers questions about our destination, which sources are describing our hotel?
04Recommended, known and referenced are three different things
Three separate signals
Recommendation ≠ known by name ≠ source reference
A hotel can be known by name yet absent from stay suggestions. Recommendation during discovery is usually harder and more valuable.
Recommendation and source representation can also differ within one answer:
Recommendation and source representation are separate measures. This proves neither causation nor clicks or bookings. Management can still examine whether its direct website is represented alongside other sources.
05ChatGPT and Gemini are different channels
ChatGPT and Gemini differ in models, search systems, data connections and update cycles. OpenAI’s web search sometimes uses partners.1 Google’s travel planning connects to Maps, Hotels and Flights.3
The same question can yield different hotel results. Examine each engine separately before combining findings.
What a hotel can actually measure
Rather than reverse-engineering AI, repeat the same questions and record five signals:
- Recommendation positionDoes the hotel appear among the options, and how prominently?
- Question coverageWhich traveller situations include the hotel: family, couples, beachfront, old town, value, wellness?
- Recognition by nameDoes the assistant describe the hotel correctly when asked about it directly?
- Source representationIs the official website among the sources, alongside booking platforms and others?
- Engine differencesDo ChatGPT and Gemini give materially different results?
Repeat measurements over time to track patterns:
What not to conclude
A website appearing as a source beside an AI answer does not prove:
- that it caused the hotel to be recommended;
- that the guest clicked it;
- that a booking happened there;
- that revenue was lost;
- that changing your website will change the next answer.
One response is not a stable ranking or “the algorithm”. Answers change over time and between sessions. Look for patterns across many questions and repeated measurements.
What this means for hotel distribution
Before rankings, sales channels and website conversion comes another distribution question:
Does the hotel enter the traveller's AI-generated shortlist at all? And if it does, which sources represent it there?
That is the layer Vektos measures.