Strategy Consulting
Years in strategy consulting to the German Mittelstand: close to the client, structured, strategic before tactical.
I make sure ChatGPT, Google AI Overviews, Perplexity, Gemini and Claude name your property when a guest asks where to stay. I work with boutique hotels, villas and small, owner-run hotel groups, based in Cologne and working internationally, and I publish the evidence behind what I claim on this page.
The difference isn’t the building. It’s whether AI has enough about you to recommend you by name.
Until recently, search returned a list. Position eight was still visible, one click away. An AI answer works differently: it names two, maybe three properties. There is no eighth position anymore. Only the answer counts, and anything missing from it stays invisible.
For that answer, AI does not rely mainly on your own website. It leans on what others write about you. In my own citation analysis, roughly four in five sources behind AI answers come from third parties, not the brand itself (own citation analysis, 2026). That is exactly where independent properties run thin: little press, few mentions, no structured presence.
The effect is measurable. Two in three hotels are never named in AI answers, even in small, easy-to-map markets (Lighthouse, 2026). This is not a problem limited to large chains. It hits the properties that should stand out the most. How strong this effect runs across markets is laid out in the data behind this page.
ChatGPT blends model knowledge with live web search for travel questions and pieces its answer together from several sources. In my own testing, what decides the outcome is less your own website than how consistently a property is described elsewhere: thin or contradictory third-party information makes it hard for the model to attach a property to a recommendation with confidence. Perplexity works differently: it shows its sources more openly and visibly favors citable material, so mentions in independent guides, trade press and best-of lists carry particular weight there (own observation, summer 2026; neither provider documents this behavior in detail).
Google confirmed in May 2026 that there are no additional technical requirements to appear in AI Overviews or AI Mode: the generative surfaces run on the same core ranking systems as classic search (Google Search Central, published May 15, 2026, updated July 10, 2026). For a hotel, that means there is no separate "AI Overview format" to maintain on top of everything else. What already ranks well and is cleanly structured in classic search holds the stronger cards here too.
Gemini leans heavily on Google's own data, including Maps and Business Profiles, so a complete, correctly maintained Google Business Profile belongs in the basics for any property, and it happens to be the cheapest item on the whole list (own observation, summer 2026).
One clarification worth having, because it is often misunderstood: Google-Extended only controls whether Gemini may use a site's content for training and grounding, not whether a property appears in AI Overviews, which Googlebot and the snippet settings control instead. Blocking Google-Extended to keep AI out loses Gemini and gains nothing in Google Search.
Anthropic split its crawler into three separate roles in February 2026: ClaudeBot for training, Claude-SearchBot for the search index, and Claude-User for requests a user triggers directly, each controllable on its own in robots.txt (Anthropic, support documentation, February 2026).
What this means in practice: a property gets cited by the retrieval crawlers, Claude-SearchBot at Anthropic, OAI-SearchBot at OpenAI, PerplexityBot at Perplexity, not by the training crawlers, which run on a separate, much slower channel. Many hotel websites block anything that looks like AI wholesale, which shuts out exactly the crawlers that make a recommendation possible in the first place.
This is not a future scenario. It is already part of how your guests decide.
Where AI does recommend, it concentrates hard. In London, 57% of all AI hotel recommendations went to just five properties (LuxDirect, 2026). If you are not one of the few names, you simply do not appear in the answer, regardless of how good your property actually is.
The demand behind this is already large. In the US market, the phrase "boutique hotel" alone appears in an estimated 16,706 AI prompts a month, with a twelve-month range of 16,700 to 28,100 (modeled estimate, DataForSEO, July 2026). Every one of those prompts decides whether your property is in the answer or nowhere at all: not page two, simply absent. The full breakdown of this estimate shows how the number is built.
This is not a threat, it is a shift already underway. It can be checked, and it can be worked on. Both start with the same question: where does your property stand today?
The starting point is a review, not a sales call. The AI Visibility Audit shows first whether and how AI names your property today, who it names instead, and why. AI Visibility Build follows: implementing what the Audit diagnosed. If you want to hold that visibility as models and competitors keep changing, AI Visibility Monitoring continues the work, entirely optional. Check where you stand, get visible, stay visible: the three steps build on each other. Scope, duration and outcome are defined for each offer. I do not name a price here.
See exactly where you stand in AI answers today.
For any independent property or brand unsure whether AI names it.
See the Audit →Go from invisible to named in the AI answer.
For properties with real character and good reviews, but thin third-party signals.
See the Build →Stay named as the models and your competitors keep changing.
For properties after AI Visibility Build that want to hold and extend their visibility.
See Monitoring →By the end, you know whether AI names your property today, who it names instead, and which signals are costing you the mention. Measurement runs against your real competitive set, not a generic comparison group.
Not included: implementation (that's AI Visibility Build) and ongoing monitoring (that's AI Visibility Monitoring).
The AI Visibility Audit is not a free hook for a sales conversation. It is a standalone product. If you commission AI Visibility Build afterward, the Audit is credited toward it.
Get visible: the Build starts where the Audit found a gap. Every diagnosis from the Audit assigns a missing mention to one of three causes, and each cause gets closed differently.
A retrieval or structure problem gets closed technically: AI crawlers are allowed in robots.txt, server-side rendering replaces pure client-side rendering, and a schema strategy plus a hub-and-spoke architecture make the site legible to retrieval systems.
A missing answer page gets closed with content. Before-and-after structural concepts become destination pages built answer-first, with data tables, Q&A structure, schema markup and clear E-E-A-T signals: pages an AI can cite, not just a guest can read.
Missing third-party authority gets closed with placement work: editorial mentions, travel-guide and best-of listings, directory and entity upkeep. This is the part no tool takes off your hands.
Evidence: a controlled study of roughly 10,000 queries shows that citations, statistics and sourced claims increase a page's share of generated answers (+41%, +31% and +30%), while keyword stuffing tends to hurt more than it helps (Princeton study, "GEO: Generative Engine Optimization," ACM SIGKDD 2024).
You receive fully written content briefs, a prioritized topic backlog, technical recommendations, an implementation checklist and a roadmap ordered by impact against effort. What you get at the end is a work order, not a slide of ideas: something you or your team can pick up directly. More on the method behind these numbers: methodology and sources.
The close is a re-audit on the same baseline: same prompts, same competitive set, so before and after are genuinely comparable.
Not included: ongoing monitoring or content afterward (that's AI Visibility Monitoring), guaranteed numbers, generic SEO.
Stay visible: Monitoring builds on the Audit, not from zero. The survey runs on the same prompts and the same structure as the Audit baseline, so real time series exist from month one instead of a new baseline with every report.
Included: a monthly survey, a short digest, a quarterly review report, and ongoing placement work: the same third-party authority building that started under AI Visibility Build, just continuous instead of one-off.
An optional add-on is a content engine running on your own infrastructure, self-hosted and GDPR-compliant, for seasonal, on-brand content without a marketing department.
The minimum commitment is six months.
Not included: tool-building, paid ads, social media management or guaranteed booking numbers.
AI Visibility Monitoring is explicitly optional, never a requirement after AI Visibility Build.
GEO is not renamed SEO: the sources AI assistants trust are partly different from the ones classic search optimization targets. It is not a dashboard that displays numbers while the signals underneath stay the same. And it is not a guarantee: no promised ranking positions, no promised booking numbers. Anyone selling you an exclusive "AI Overview special format" is selling you something that, by Google's own account, is not required.
That is Google's own position, stated in its documentation and quoted above. It has a consequence worth spelling out, including for what I sell: the work here is largely the same work that makes a site strong in classic search, done for a different reader. There is no proprietary trick, and anyone offering you one is describing something that does not exist.
The claims on this page rest on four kinds of sources: dated industry surveys (Booking.com 2025, Lighthouse 2026, LuxDirect 2026, DataForSEO July 2026), the only peer-reviewed controlled study on generative search optimization to date (Princeton University et al., ACM SIGKDD 2024, roughly 10,000 queries), Google's own documentation on AI Overviews and AI Mode (Google Search Central, published May 15, 2026, updated July 10, 2026), Anthropic's own crawler documentation (support documentation, February 2026), and my own measurements with a stated survey date (citation analysis and prompt testing, summer 2026).
What these sources do not provide: a guarantee for your specific property. The Princeton study measures share of generated answers, not traffic or bookings. Share-of-voice figures from my own surveys are directional, not an exact measurement. None of these sources say anything about your hotel before I have actually checked it. Full sources and survey method: the research page.
This work did not start with a founding idea. It started with direct experience. Years in strategy consulting to the German Mittelstand: close to the client, structured, strategic before tactical. Building CITRA Villas in Lombok from company formation to brand, and at the same time my own testing ground for visibility work. Supporting Hijack Sandals' entry into the European market, as proof of craft, not as an audience signal.
Years in strategy consulting to the German Mittelstand: close to the client, structured, strategic before tactical.
Building CITRA Villas in Lombok from company formation to brand, and at the same time my own testing ground for visibility work.
Visit CITRA ↗Supporting Hijack Sandals’ entry into the European market, as proof of craft, not as an audience signal.
See Hijack Sandals ↗Independent, owner-run stays that win on character, not scale, and deserve to be named for it.
A handful of properties that need to show up consistently, not just the flagship.
Independent, design-led properties with a voice of their own: often exactly the kind AI flattens into generic language instead of naming.
This focus has a reason, and it looks different in English than it does in German. In the English-speaking market, hospitality-specific AI-visibility providers already exist. What most of them sell is a tool or a dashboard. I have built and run an actual hospitality brand myself, CITRA Villas in Lombok, and I still operate it. That is the difference: not a missing category, but a different way of doing the work inside it.
My own research on generative search in hospitality is the foundation of this work. In English, it is published as "The Disappearing Hotel." It draws on my own prompt testing and citation analysis to show that independent hotels appear structurally less often in AI answers than chain properties, not because they are worse, but because they lack the third-party sources models rely on. The full version, with methodology and raw data: The Disappearing Hotel.
Request the Audit. Reachable by email, WhatsApp or LinkedIn.
Email: hello@victorfrenzel.com · WhatsApp: Send a message · LinkedIn: linkedin.com/in/victor-frenzel