AI Overviews Are Eating Booking.com's Traffic, and What Hotels Should Do Now
AI assistants are answering guest questions without sending the click to Booking.com or your website. Here is what that means for hotels, and the four moves worth making this quarter
A guest opens ChatGPT on a Tuesday evening and types, "Find me a romantic boutique hotel in Charleston for next weekend, with a clawfoot bathtub and a quiet room away from King Street." Thirty seconds later, she has three recommendations, a short paragraph on each, and a sense of which one is right for her anniversary. She never opened Booking.com. She never typed your hotel's name into Google. She may, if you are lucky, click through to your website to finish the booking. More likely, she will book directly with whichever property the assistant placed at the top of its answer, on whatever channel the assistant suggested.
This is the quiet shift happening underneath the hotel industry right now. It is not a marketing trend. It is a distribution event, and it is moving faster than most operators realize.
The traffic is already moving
Anyone who has watched their Google Analytics dashboard over the last twelve months has seen the pattern. Organic traffic to hotel websites has softened. OTA referral volume has flattened or declined in some segments. Booking.com's own results have hinted at slower growth in the search-driven middle of the funnel, and they are not alone. Industry data suggests AI Overviews, the answers Google now generates above the traditional ten blue links, are absorbing a meaningful share of the clicks that used to flow downstream to travel sites.
The mechanism is simple. When a question can be answered inside the search result, the click never happens. The user gets what they came for, the answer, and the website that supplied the underlying information often receives nothing. No referral. No session. No remarketing pixel. No booking.
Reports indicate this is happening across categories. Hotels are particularly exposed because travel research has always been a high-effort, multi-tab activity. Compressing it into a single conversational answer is exactly the kind of task that AI assistants are good at, and exactly the kind of task that consumers want compressed.
The question is no longer how to rank on Booking.com. The question is whether the AI answering your guest's question even knows you exist.
What the AI is actually doing
When a guest asks Claude, ChatGPT, Perplexity, or Gemini for a hotel recommendation, the assistant is doing something quite different from a Google search. It is reading whatever sources it can access (some live, some from training data, some via tool calls to web search or specialized servers), then synthesizing an opinion. That opinion is shaped by three things: what data the assistant can see about your property, how recent that data is, and whether the assistant trusts it.
A hotel with a well-maintained Google business profile, a sharp website, accurate listings on a few high-trust travel sources, and a description that matches what a guest is actually asking for, has a chance of showing up. A hotel with stale photos, a thin website, no structured data about room types, and an old description that refers to a restaurant that closed in 2023, has very little chance.
The clawfoot-bathtub guest asking about Charleston is a perfect case study. The assistant will look for properties described as romantic, boutique, in Charleston, with specific room features. If the only place that information lives is in a few TripAdvisor reviews from four years ago, your hotel is invisible to that conversation. If the information lives in a structured, machine-readable place that the assistant can query directly, you are in the running.
What hotels should do now
There are four moves worth making this quarter. None of them require firing your marketing agency or rebuilding your website. They require getting your data house in order before the rest of the industry catches up.
1. Get your room-level data out of the OTAs and into your own control
Booking.com has your inventory. Expedia has your inventory. You have, in most cases, a homepage and a booking widget. That asymmetry is a problem when AI assistants start asking specific questions about which of your rooms has a soaking tub, which is on the courtyard side, which is ADA accessible, and which one is the corner suite with two windows.
Your room-level data needs to live somewhere you control, in a format that machines can read. This is the single highest-leverage change a hotel can make in 2026. The hotels we work with through Roomza have been moving in this direction for two years, and the gap between properties with rich, structured room data and those without is already visible in how AI assistants describe them.
2. Stand up an MCP server
Model Context Protocol is the emerging standard for how AI agents query structured data. A hotel with an MCP endpoint can be asked, in real time, by Claude or any compatible assistant: "What rooms do you have available that are quiet, have a clawfoot tub, and are under $400 a night?" The assistant gets a precise answer. The hotel gets a precise referral.
You do not need an engineering team to do this. Roomza ships an MCP server for every hotel on the platform. If you want to understand the mechanics first, we wrote a primer: Why Your Hotel Needs an MCP Server, and How to Do It for Free (or Not Overpay).
3. Write a concierge brief
This is the part that hoteliers often miss. AI assistants benefit from a short, well-written, opinionated summary of your property. Not marketing copy. A brief that tells the assistant who this hotel is for, what makes it different, what to recommend it for, and what to warn guests about. A two-paragraph honest brief beats a six-page brochure every time.
The Roomza concierge brief is editable by the hotel. When a guest asks an AI assistant about your property, that brief is one of the things the assistant can read. You get to shape, in your own words, the answer the AI gives.
4. Get serious about direct distribution
If AI assistants begin pointing guests at the cheapest available channel, and your cheapest channel is your own website, the math starts to favor direct bookings in a way it has not in a decade. The hotels that win the next five years will be the ones that quietly built their direct channel, kept rate parity honest, and gave AI assistants a clear reason to recommend the direct booking link.
A once-in-twenty-years shift
There have been three real distribution events in modern hotel history. The arrival of online travel agencies in the late 1990s. The mobile booking shift around 2012. And now this. Each of those moments rewarded the operators who moved first and punished the ones who waited to see how it played out.
The hotels we are watching closely right now are not the chains. They are the independent and small-group operators who are reading this trend correctly: that AI-driven discovery is going to be the dominant top-of-funnel channel within a few years, that the structured-data layer underneath it is being built right now, and that the cost of getting on the right side of this shift is far lower today than it will be in eighteen months.
You do not need to predict exactly how this plays out. You need to make sure the answer the AI gives your next guest is the answer you want it to give. That starts with owning your data, exposing it cleanly, and writing a brief that sounds like the hotel you actually run.
The clawfoot bathtub is waiting. The question is whether the assistant knows where to find it.
