CITRA Villas
Building CITRA Villas on Lombok from company formation to brand, and at the same time my own testing ground for the visibility work.
Visit CITRA ↗Does ChatGPT Know Your Hotel?
One question to four AI assistants. Each answers differently, and the same way on two days.
I make sure AI assistants such as ChatGPT name your property when a guest asks for a hotel. For boutique hotels, villas and small hotel groups, from Cologne.
Tested for Haus im Tal, Munich, not a client. The question, verbatim:
“Design hotel in Munich’s old town with a roof terrace and its own bar, not a chain.” Original„Designhotel in der Münchner Altstadt mit Dachterrasse und eigener Bar, kein Kettenbetrieb“
| Google AI Mode | top recommendation “The perfect hotel that meets all your criteria exactly is the ANAGRAM - Haus im Tal.” Original„Das perfekte Hotel, das exakt alle Ihre Kriterien erfüllt, ist das ANAGRAM - Haus im Tal.“ |
|---|---|
| Google AI Overview | no AI answer Same company, same question. For this wording Google served ads and a hotel pack instead. |
| Perplexity | recommends a different property “A design hotel in Munich’s old town with its own roof terrace and bar that is not a classic chain operation is the DO & CO Hotel München.” In the same paragraph Perplexity calls it “part of the Austrian DO & CO group”. Original„Ein Designhotel in der Münchner Altstadt, das eine eigene Dachterrasse und Bar hat und kein klassischer Kettenbetrieb ist, ist das DO & CO Hotel München.“ |
| ChatGPT | named in second place The closing recommendation goes to the LOUIS Hotel. Being named and being recommended are not the same thing. |
- Surveyed 24.08.2026, 17:45 to 18:18 CEST, and 25.08.2026, 10:31 to 12:31 CEST. Asked in German, Munich market.
- Two runs per surface, the same verdict in all four. A third to follow.
- Models do not answer deterministically. A repeat run can turn out differently.
The shift
Search just changed its front door
Google returned a list, and position eight was one click away. An AI names two or three properties. Whoever is not among them is invisible.
Four in five sources an AI answer rests on come from third parties (own analysis of the cited sources, 2026). Hardly anyone writes about independent properties.
It hits the properties that differ most from the crowd. The international figures in detail.
The consequences
What this means for your property
An AI answer names two or three properties. In London, five hotels share more than half of all recommendations. Whoever is not among them does not appear, however good the property is.
This shift is already underway. It can be checked, and it can be worked on. Both begin with the same question: where does your property stand today?
67%
of travelers already use AI somewhere in planning a trip
Booking.com, 2025
2 in 3
hotels are never named in AI answers, even in small markets
Lighthouse, 2026
57%
of London’s AI hotel picks went to just five hotels
LuxDirect, 2026
All sources on this page, with dates: the reference list of the study.
The path
Five steps, the first two free
The first two steps cost nothing. The Self-Check tells you in seconds whether an AI can read your site. The Results tell you within 24 hours whether it names your property. The Hotel Visibility Audit says why not and what to do. The Build implements it, Monitoring keeps measuring.
Step 0a · free, instant
The Self-Check
Tells you in seconds whether an AI can read your site. By email: a score from 0 to 100, every check with its points.
Start the Self-Check →
Step 0b · free, 24 hours
The Results
Three questions to an AI, asked the way a guest asks them. The answers verbatim, and who gets named in them.
Tick it in the Self-Check →
Step 1 · ten working days
The Hotel Visibility Audit
25 questions to ChatGPT, Perplexity and Gemini, the main cause behind the answers, and 8 to 12 measures with ready-to-use text.
See the Audit →
Step 2 · after the Audit
AI Visibility Build
Starts where the Audit found a gap and closes it by its cause: technical, editorial, through third-party sources or through directories. I credit the Audit in full.
Roughly ten to twelve weeks. Ends with a re-audit on the same baseline.
Step 3 · monthly
AI Visibility Monitoring
Asks the same questions every month, so you see what moves.
Minimum six months. A one-page digest, a quarterly call.
Build and Monitoring follow the Audit. Their scope comes in a written proposal once the Audit is on the table.
Step 1 · The Hotel Visibility Audit
How do you find out whether the AI names your hotel?
25 guest questions to ChatGPT, Perplexity and Gemini, the cause, 8 to 12 measures. In ten working days, ordered without a call.
By the end, you know three things
Whether the AI names your property today, whom it names instead, and why. Mostly it comes down to what others write about you. Your own website is rarely the bottleneck. How robust that is across markets, the study shows in detail.
How I proceed
- Survey. 25 questions, asked the way guests ask them for a property like yours, in five groups: by name, by description without the name, the category question, the question before the hotel question, and the direct comparison with your competitors. Each goes once to ChatGPT, Perplexity and Gemini, with date and model version per answer.
- Competitors and sources. The same questions run against the properties the AI names. You see who gets named, how often and where, and which pages the AI reads for it: trade press, review portals, guides, forums, directories. And which factual errors the AI spreads about your property, verbatim.
- Diagnosis. Every gap gets a cause, and the cause decides what helps: the systems do not read your site cleanly, no page answers the question, the AI cites others and you are not there, or the AI confuses your property. The usual case in this industry is the third.
- Measures. 8 to 12 measures in the order in which they work, each with the gap it follows from, rated by impact against effort, with ready-to-use text.
What you receive
- The Audit as a PDF, 10 to 14 pages, with a summary up front
- The question set and the full raw data of the survey, with a source register
- 8 to 12 prioritized measures with ready-to-use text
- The metrics with their starting values for later measurement
- 30 minutes of follow-up questions by phone
What happens on your website shows within weeks. Mentions take months. I say so beforehand. If you commission the Build afterwards, I credit the Audit in full.
The order
Order the Hotel Visibility Audit
Without a call. Within 24 hours I look at your website and send you the order confirmation with scope and price; it takes effect as soon as you reply with one word. For 48 hours after that, either of us can withdraw without giving a reason. The invoice follows, payable within 14 days; the survey starts with the invoice.
Track record
A consultant who built all of this himself
I built a property on Lombok myself and then checked whether ChatGPT knew it. It did not. This work grew out of that. Before that, years of strategy consulting for German mid-sized companies: close to the client, structured, the strategy before the tactics.
Who it’s for
Built for places worth recommending
Boutique hotels & villas
Independent properties with a design ambition that win on character, not scale.
Small, owner-run hotel groups
A handful of properties that want to stay visible consistently, everywhere.
I work only with hotels, because a hotel has different questions, different competitors and different sources than a mid-sized company.
The research
The whole picture, written down
Why independent hotels appear in AI answers less often than chains, with the international figures and my own citation analysis. About 20 minutes of reading, also as a PDF. Two further papers, on structured data and on keyword research, are in German: all research pages.
FAQ
What most people ask first
Is this real, or AI hype?
Yes, it is measurable. The panel at the top of this page is a measurement from August 2026. Whether the AI names your property, the Results show you, free within 24 hours.
Can you measure this?
Yes, with one caveat: the same question can be answered differently tomorrow. That is why I always ask the same 25 questions and compare over time, not a single day.
Why you, and not an agency or a tool?
When I asked an assistant in July 2026, from the perspective of a German boutique hotelier, how to get into AI recommendations, five of five source citations came from Google’s own documentation (own test, July 2026). Advice content does not get you further. A tool shows you numbers, but it does not call an editor. The work on the third-party sources I do myself.
Who owns the content that the Build produces?
You do. Everything created in the Build appears on your domain under your name and stays yours.
The first step costs nothing and takes seconds.
The Self-Check reads your homepage in seconds. Tick the free Results there, and within 24 hours you see whom the AI names instead of you.
Email: hello@victorfrenzel.com · I usually reply within 24 hours.