When someone asks AI where to stay, does it name you or Booking.com?

What does an AI read when someone asks where to stay?

Mostly other people's listings of you. When a guest types "where should I camp near St Ives with a dog" into ChatGPT, the assistant does not know your site. It reads pages about you, and the pages it reaches for first are the ones that carry prices, dates and reviews in a tidy, repeatable shape. That describes a booking platform far better than it describes most campsite websites.

Cloudbeds, which sells software to hotels, ran 810 prompts through ChatGPT, Perplexity and Gemini across six destinations and looked at the 145 properties that kept coming back. Of everything those three assistants quoted, 55.3% came from online travel agents such as Tripadvisor, Booking.com and Expedia, and 13.6% came from the property's own website.1 Travel blogs were 19.2%, forums including Reddit were 6%, and video reviews 1.6%.

Where the answer comes from

What AI assistants quoted when recommending a place to stay

Booking platforms 55.3%

Travel blogs 19.2%

The property's own website 13.6%

Forums, including Reddit 6%

Video reviews 1.6%

The three assistants did not behave the same way. Cloudbeds found Gemini leaned heavily on Booking.com, Perplexity leaned on Tripadvisor, and ChatGPT referred to properties' own websites more often than the other two did.1 That last point matters, because ChatGPT is the assistant most of your UK guests are using.

One number in that study is worth holding on to for later: 13.6% is described by Cloudbeds as unusually high, against a figure of about 9% for company websites across other industries. Accommodation businesses are actually doing better than most at being quoted from their own site. It is still one source in seven.

Does it name independent businesses at all?

Yes, but a small and lopsided share of the time. Lighthouse, which sells commercial software to hotels, fed 4,545 prompts into ChatGPT across nine destinations in the United States, Europe and Asia, using five different types of traveller. Presenting the results in June 2026, it reported that ChatGPT named hotels 49,707 times, but those mentions covered only 2,721 different properties.2

The same study measured how much of each local market ever got mentioned. In Tokyo it was 10% of hotels, in Paris 13%, and in Park City, a small resort town, 33%. Two thirds of the properties in that small town were never named once across 101 different ways of asking.2

Independent properties took the thin end of what was left. In Indianapolis, 6.5% of hotel mentions went to independents. In Tokyo it was 10.5%. In Brussels it was 28%, despite the city having more independent hotels than chain ones. Paris was the only market Lighthouse tested where independents were named more often than chains.2

Who gets named

Share of hotel mentions that went to independent properties

Indianapolis 6.5%

Tokyo 10.5%

Brussels 28%

Neither study looked at a UK campsite, a glamping field or a holiday cottage. Both were run by companies with something to sell to the businesses they were studying. Take the direction of travel, not the decimal places.

How many UK customers are actually asking?

Not many, yet, and you should be wary of anyone who tells you otherwise. Ofcom's Online Nation report, published on 10 December 2025, found that ChatGPT reached 13% of UK online adults in May 2025, up from 9% a year earlier, while Google search reached 82%.3 Nine in ten of your guests are still starting somewhere other than an AI assistant.

The bigger change is the one nobody chose. Around 30% of keyword searches now return an AI-written summary above the results, and Ofcom reports research finding that 53% of UK people say they often see those summaries.3 Your customers do not have to adopt anything for this to reach them. It arrives in the search they were already doing.

What happens next is measurable, and it is not good for whoever was hoping for the click. Pew Research Center installed a browsing tracker on the computers of 900 American adults and recorded 68,879 Google searches during March 2025. When an AI summary appeared, people clicked through to a website on 8% of visits, against 15% of visits with no summary. They clicked a source listed inside the summary on 1% of visits.4

That study is American and covers Google only, so read it as a shape rather than a UK forecast. The shape is this: being mentioned in the answer is worth more than it used to be, and being merely listed underneath it is worth less.

Why does it matter which name gets said?

Because the two answers cost you different amounts of money. If an assistant recommends your site and the guest reaches you directly, the booking is yours. If it recommends your listing on a platform, the same guest, the same night, the same pitch, now arrives with a percentage attached.

Pitchup.com's supplier FAQ stated a 15% commission on bookings it sends, when we checked it on 24 August 2026.5 Booking.com does not publish a rate at all. Its partner help says the exact percentage "depends on your country, property type and the accommodation agreement you signed", and that optional programmes such as Preferred Partner and Visibility Booster add a further percentage on top of your base rate.6 If you cannot see the rate written down anywhere public, that is not an oversight.

The route from question to booking is also getting shorter, in the platforms' favour. Since 8 October 2025, Booking.com and Expedia have run as apps inside ChatGPT, showing live prices, photographs and availability in the conversation. A guest can call the app by name, or ChatGPT can offer it. The reservation itself is completed on the platform's own site.7

The path a guest takes

How a recommendation turns into someone else's booking

01

They ask

"Somewhere to camp near Padstow in half term, with hook-up and a dog."

02

It reads what it can

Listings, review sites and travel blogs carry prices and dates in a shape it can use.

03

It names a shortlist

Your site may be on it. The link beside your name is often a platform's.

04

They book there

Same guest, same pitch, now with a commission attached to it.

None of this makes the platforms villains. They earn that commission by putting your pitches in front of people who had never heard of you. The problem is the guest who already knows your name, has been recommended you by an assistant, and still ends up booking through a middleman, because the middleman's page was the one with the price on it.

What do UK campsite websites give an assistant to work with?

Often very little. There is no published research on this for UK campsites, so on 25 August 2026 we ran our own check. We took every campsite, caravan park and holiday park that records a website in OpenStreetMap across four UK holiday areas, Cornwall and Devon, the Lake District, North Yorkshire and Pembrokeshire, which gave us 496 separate businesses. We then visited each website the way an automated reader would, looking at the homepage and up to two pages linked from it whose wording suggested prices or booking.9

90 of the 496 sites did not load for us at all. Some had expired certificates, some returned an error, a few were near empty. That is 18% of these businesses invisible to an automated visitor on the day we looked, before any question of what the site said.

Of the 406 that did load, 193, or 47.5%, showed a price. Just over half offered any way to book or check availability on their own website. Roughly a quarter showed neither a price nor a booking step, which leaves a guest who wants to book with you directly nothing to do but ring in the morning.

Frontpage AI research

What 406 working UK campsite and holiday park websites showed us

Counts and percentages for five checks run against 406 reachable UK campsite and holiday park websites on 25 August 2026.

What we looked for	Sites	Share
A postcode anywhere on the page	314	77.3%
A way to book or check dates on their own site	233	57.4%
Any hidden code spelling out their facts for machines	218	53.7%
A price	193	47.5%
Hidden code that says what kind of business this is	63	15.5%

One sample, on one day, using blunt automated tests, is not a study of the whole UK, and we are not presenting it as one. A site that keeps its prices on a separate tariff page two clicks in will show up here as having no price, which is unfair to the owner and completely fair to the assistant, because the assistant does not dig either.

Set that beside the Cloudbeds finding and the picture joins up. Assistants quote booking platforms most of the time. Booking platforms publish prices, dates and availability for every property on them, every day. Half of these businesses publish no price at all on the pages a reader lands on first. The platform is not winning the argument. It is the only one making it.

What can you actually change?

Less than the tools industry implies, and more than nothing. Start with what Google says, because Google is the only one of these companies that publishes its rules. Its documentation, updated on 10 December 2025, says a page must be indexed and eligible to be shown in ordinary Google search with a snippet, meaning it already appears in normal results and Google is allowed to show a line or two of its text. Google states there are no additional technical requirements, and that you do not need new machine-readable files or special markup to appear in AI Overviews or AI Mode.8

So the work is unglamorous, and most of it you can do yourself this week. Put your prices on your website, in text, with the season and what is included. Put your postcode and your nearest town in words, not only on a map that an automated reader cannot see. Say what a pitch actually has: hook-up, hardstanding, dogs, showers, the walk to the beach in minutes. Let people book, or at least check dates, without leaving your site. Keep asking for reviews on the platforms you are already on, because those pages are what assistants read most.

You can do all of that without hiring anybody, and if you have the time, you should. What people usually hire us for is that it stays done: prices updated when the season changes, the site loading for every automated visitor rather than 82% of them, the facts consistent between your website, your Google listing and your platform listings.

Frontpage AI is a UK studio that builds and manages websites for independent service businesses, structured so that Google and AI answer engines can read, trust and name them. We work only with UK service businesses, and charge from £88 per month on a 30-day rolling contract with no setup fee. For campsites, glamping sites and others that take bookings through a platform, we work on 5% of the bookings your own site takes instead of a monthly fee, because a site that takes nothing in February should pay nothing in February.

If you want the reasoning behind all of this rather than the sales version, the how we approach AI search page sets it out, and our article on whether a paid directory listing is still worth it covers the same question for trades.

What this evidence does not tell you

It does not tell you what an assistant says about your campsite. Every study here is about hotels, and mostly hotels in American, European and Asian cities. A hotel in Tokyo and a farm with twelve pitches outside Hawes have almost nothing in common except a bed.

It does not tell you how much money is moving. There is no published figure for how many UK camping bookings begin with an AI assistant, and we could not find one that survived being checked. Anyone quoting you a percentage on that is guessing.

Both hotel studies were published by companies selling software to the businesses in them. That does not make the numbers wrong, and both describe their method openly, which is more than most of the AI visibility research on the internet does. It does mean neither had a commercial reason to conclude that this is nothing to worry about.

And our own check measures what an automated reader can see in a few seconds, not how good these businesses are. Some of the sites with no price on them are lovely places that fill every August without help from anyone. The question this article asks is narrower: when a machine goes looking for a reason to name you, is there one on your website. For about half of them, on the day we looked, there was not.