Strategy & Planning · 9 min read · July 15, 2026
Away from commission: how GEO wins back direct bookings for hotels
Every booking through Booking.com or Expedia costs you 15 to 25 percent commission. But more and more guests no longer ask the portal – they ask ChatGPT, Gemini or Google AI Overviews: "Where do I stay in the Allgäu with a dog and a sauna?" Whoever appears there as a recommendation wins the guest before the portal even comes into play. That is exactly what GEO delivers.
The commission problem every hotelier knows
Run the numbers honestly once. At a room rate of 140 euros per night and a Booking commission of 18 percent, you lose a good 25 euros per night. Over a three-night stay that's 75 euros that don't stay with you but go to the portal. Across a year with a high OTA share, we quickly reach a five-figure sum that goes entirely to margin – money you could have put into renovation, staff or breakfast quality.
The truly bitter part isn't the commission itself, but that for many of these guests you pay twice. They find you via the portal, book there, and next time they book there again, because the portal owns the customer relationship and you don't. The guest knows Booking, not your name. So your task isn't only to get bookings, but to win back the relationship with the guest before the portal captures it.
This is exactly where the first point of contact is currently shifting. In the past the guest typed a search into Google, saw a Booking ad and was in the portal's funnel. Today the same guest asks an AI their question and gets a concrete recommendation by name. Whoever occupies this new first point of contact bypasses the portal funnel entirely.
What GEO is and why it works differently for hotels
GEO stands for Generative Engine Optimization – optimising your visibility in AI systems such as ChatGPT, Google Gemini, Perplexity and the AI Overviews in Google Search. Unlike classic SEO, it isn't about ranking first on a results list. It's about the AI knowing your property at all, classifying it correctly and naming it as an answer at the right moment. There are no longer ten blue links, but often just a handful of named properties.
For hotels this is especially relevant, because travel decisions are extremely context-rich. A guest rarely asks just for a hotel in Munich. They ask for a quiet hotel in Munich with parking, near the trade fair, dog-friendly and with late check-in. The AI combines all these criteria and needs properties whose information supplies exactly these details cleanly and machine-readably. Whoever states none of these details clearly won't appear in the combination.
The decisive difference from the portal: the AI draws its answers from the entire open web – from your website, from reviews, from trade directories and from press articles. So you no longer depend on how highly a portal places you in its internal sorting. You can actively shape the sources the AI learns from.
How guests really search for hotels today
The language of AI search is spoken, everyday language. Instead of "wellness hotel Black Forest", the guest types: We're a couple in our late thirties, we want to get away for a weekend in the Black Forest, good food, a sauna, but not that over-the-top luxury. What do you recommend? The AI answers with two to four specific properties and a short reason why they fit. That very reasoning decides whether your property is named or not.
Think of the typical questions of your real guests. Is breakfast ready for early risers before the hike? Can I charge my electric car? Is it really quiet, or does it sit on a through road? Do you have family rooms for four people? Each of these questions is a potential AI query. If your website and your other sources answer them clearly, the AI has material to recommend you.
The problem with most hotel sites is that they wallow in marketing language instead of facts. A sentence like "Enjoy moments of deceleration in our retreat" tells an AI nothing. The sentence "All rooms face the quiet garden side, breakfast is served from 6:30 a.m., charging stations for electric cars are available free of charge", by contrast, delivers exactly the facts that carry a recommendation.
Your website as a machine-readable source of facts
The most important lever is your own website, because it is the only source you fully control. Make sure all the hard facts are stated in clear text and not hidden away in an image or a PDF. Location with concrete distance to the station, ski lift or trade fair, room categories with square metres and occupancy, amenities, check-in times, pet policy, parking situation and price ranges belong on the page as readable text.
Add structured data in the Schema.org format, specifically the Hotel or LodgingBusiness type. This gives search engines and AI systems your facts in a format they are guaranteed to read correctly. Address, star rating, price range, amenity features and reviews can be stored unambiguously this way. It's invisible to the guest but a strong signal for the machine, and it reduces the risk of the AI misclassifying your property.
Also build an honest FAQ section that answers exactly your guests' questions in their own language. This question-and-answer structure is gold for AI systems, because it matches precisely the pattern the AI itself thinks in. A good hotel FAQ today is no longer a service extra but a central instrument of your AI visibility.
Reviews and external sources as a trust signal
AI systems trust your own website only to a limited degree, because by its nature it carries advertising. So they cross-check your statements against external sources, above all guest reviews. If you speak of a quiet location on your site and the word "street noise" keeps appearing in the reviews, the AI won't recommend you for that context. Reviews are thus not just reputation but direct GEO input.
Work actively to get the right, concrete terms into reviews. When your staff ask at check-out what the guest especially liked, and the guest mentions it in their review, exactly the phrasings arise that an AI will later pick up. A guest who writes "Perfect for hiking trips, breakfast as early as six" makes your property recommendable for an entire category of enquiries.
Also mind consistency across all directories. Name, address and phone number should be identical in Google Business, on travel portals and in regional directories. Contradictory information confuses the AI and weakens its trust in your data. A clean, consistent digital footprint is the foundation on which every AI recommendation builds.
Occupy niches instead of charging at the big players
Against the query "hotel in Berlin" you as a single property will struggle to break through in the AI, because the competition is too large and too unspecific. Your chance lies in the sharp, concrete queries. Bike-friendly hotel on the Danube route with a lockable bike cellar. Family hotel in the Allgäu with a pool and childcare in the afternoon. Hotel in Hamburg for business travellers with quiet rooms and fast Wi-Fi.
For these specific combinations there are far fewer matching properties, and if you highlight your special features clearly, you are often one of only two or three named. Consider which three to five characteristics truly make your property unique, and weave these consistently into your entire communication, from the website to the FAQ to the replies to reviews.
This niche strategy has a pleasant side effect. Guests who come via a very precise recommendation are more satisfied, cancel less often and fit your offering better. So you win not only commission-free bookings but the right guests for your property.
From AI contact to direct booking
Visibility in the AI only helps if the path to direct booking afterwards is smooth. When a guest discovers your property via ChatGPT and opens your website, a simple, fast booking route must be visible there immediately. An overloaded homepage, a hidden booking button or a slow-loading booking form send the guest back to the portal, where everything is convenient. The last few metres decide over the commission.
Give guests an honest reason to book directly instead of through the portal. A guaranteed best price, a free late check-out, a glass of wine on arrival or flexible cancellation only for direct bookings are strong arguments. Communicate this advantage clearly and prominently, because in that moment the guest is actively comparing with the portal price they have in mind.
Also measure whether your work is having an effect. Regularly ask the AI systems the typical guest questions of your region yourself and observe whether and how your property is named. This simple check shows you in black and white where you are already visible and where gaps still yawn that you can work on in a targeted way.
A realistic roadmap for the coming months
Don't start with everything at once, but with the biggest lever. In the first step you get your website in order, make all facts readable, add structured data and build an honest FAQ. That alone already puts you in play in many AI answers and is the foundation for everything else. Plan a few weeks of focused work for this, ideally with expert support for the technical parts.
In the second step you take care of the external signals, that is, systematic review management and the consistency of your data across all directories. In the third step you sharpen your niche positioning and optimise the direct booking route with clear advantages. This triad is not a one-off action but an ongoing process that you review every few months.
Be honest about your expectations. GEO won't replace the portals overnight, and for short-term occupancy they remain a tool. But every booking you win directly through your AI visibility is a booking without commission and with a customer relationship that belongs to you. Over a year that adds up to a real difference in your margin and your independence.
Common questions
Will I lose my visibility on Booking if I focus more on direct bookings?
No. GEO and portals aren't mutually exclusive. The portals sort by their own criteria, regardless of whether you are additionally visible in AI systems. A coexistence makes sense: the portals bring you reach and occupancy in weak periods, while your AI visibility and the direct booking route win back commission-free bookings and the customer relationship. Just make sure not to set your direct price below the portal price, to avoid rate-parity conflicts, and put the added value into extras like late check-out instead.
We're a small family hotel without our own marketing department. Is GEO even worth it for us?
For small properties in particular it is often more worthwhile than for large chains. Your strength is sharp, concrete features like dog-friendly, quiet location or family rooms with childcare. There is little competition for exactly such specific queries, so you get named as one of only two or three properties. Getting started is manageable: a clearly structured website with readable facts, an honest FAQ and active review management. This can be implemented step by step, without a big budget, and works directly toward commission-free bookings.
How do I know whether my hotel is already being named in AI systems like ChatGPT?
Test it yourself, regularly and systematically. Ask ChatGPT, Gemini and Perplexity the typical questions of your guests in everyday language, for example a quiet hotel in your town with parking and a sauna. Note whether your property is named, in what position and with what reasoning. Also check whether the stated facts are correct or whether the AI is misclassifying you. This short check every few weeks shows you exactly where you are already visible and which features or sources you still need to work on.
Read on