What ChatGPT sees when someone asks for places to stay near you

A family is planning a week in Cornwall. They're not opening Booking.com. They're not scrolling through Instagram. They're typing into ChatGPT: "We need a dog-friendly cottage near Padstow for half term, somewhere with a hot tub, sleeps six." In under a second they get a shortlist of five cottages with names, locations, and reasons why each one fits.

Are you on that shortlist?

If you're an independent operator, ten to thirty units, no marketing team, a website your web person built five or six years ago, I can almost guarantee you've never checked. Not because you don't care, but because nobody has told you this is a thing you need to check. Your Google Analytics looks fine. Your direct bookings are holding up. The numbers on your spreadsheet haven't moved much.

But the traffic that matters is moving somewhere you can't see.

This wasn't a problem eighteen months ago

I've been working with hospitality businesses for two decades. I've built websites, helped operators become bookable online, watched the shift from brochure sites to booking engines to social-first marketing. Every few years there's a new thing to worry about, and most of them turn out to be moderate nudges rather than seismic shifts.

This one is different, and the scale of the shift is what makes me want to tell every operator I know.

In March 2026, Phocuswright published research showing that 56% of US leisure travellers used AI tools for at least one trip in the past twelve months. Not "might use in future". Already have. In the UK the numbers are catching up fast. That's the fastest behavioural shift travel has seen in a decade, faster than smartphones, faster than social media, faster than the move to mobile booking.

And if you add one number to that, it gets uncomfortable: 84% of travellers say a trusted AI recommendation would make them more likely to book a specific property. That's from a TakeUp study published in January 2026. Trust is moving. The thing guests used to trust, the Google results page, the reviews pile, the word-of-mouth tip from a friend, is being partly replaced by a conversation with a machine.

Why the big hotels are already moving

Hyatt told its investors in February 2026 that its new AI-driven search system is producing stronger booking conversion, higher revenue per booking, and longer average length of stay. Multiple quarters of data. Hilton, Marriott, Accor, all either inside the ChatGPT app directory already or rolling out natural language search on their own sites during 2026.

The chains have teams for this. Independents don't. That's the gap glassmark exists to close.

What a machine actually sees

Here's the thing nobody tells you. When ChatGPT visits your website to work out whether to recommend you, it doesn't see what a human visitor sees. It doesn't see your beautiful photography, your lovely fonts, your careful colour palette, or the thoughtful copy you wrote about the view from the hill.

It sees labels.

Think of schema like the label sewn into a jumper. The label tells a machine at a glance: size medium, 100% wool, made in Portugal, machine wash cold. Without the label, a machine has to guess. It can see there's a jumper, probably, but it can't confidently say what size, what material, or where to send it. A human can pick the jumper up, feel the weight of the wool, check the stitching. A machine needs the label.

Your website has two audiences now. Humans, who will look at your photos and read your words and feel the place. And machines, who need clear structured labels telling them what kind of business you are, where you're located, what you offer, who it's for, and how to book. Most independent hospitality websites have the first audience brilliantly handled. Very few have thought about the second.

glassmark scans your site and shows you what the machines can and can't read. It's not subjective. It's not a marketing opinion. It's a measurement of whether the labels are there.

Why the problem is only going to get worse

The speed of the shift is the thing. Language models ship new versions on roughly monthly cycles. Each release reshuffles which sources it trusts and which signals it weights most. A study of how ChatGPT cited hospitality sites three months ago is already partly out of date because two of the models it studied have been retired and replaced.

Chasing individual model quirks is a losing game. What wins is making your website so structurally clear and so factually consistent that any machine reads you correctly the first time. That means consistency across every surface your business appears on: your site, your Google Business Profile, your OTA listings, your review pages. Any machine that exists today, any machine that ships tomorrow, and the one nobody has built yet.

Put the labels in properly and every current and future machine can read them. Leave them out and you're invisible to each one, regardless of which model is in fashion this month.

What you can do this week

First, see where you actually stand. Run a free scan at glassmark.ai. It takes under a minute, no sign-up, no sales call, no credit card. You'll see your overall score, six category breakdowns, and your top three priority actions. If the score is good, fantastic, you don't need me. If the score is hidden or partial, you know exactly where to start.

Second, keep doing the things you were already doing. Writing your blog, taking beautiful photos, being thoughtful about your brand. Those are exactly the signals that make your business distinctive to humans. The work glassmark highlights is the translation layer, making sure the machines can read what humans already love.

Third, don't panic. The structural improvements we recommend are the same improvements that drive better performance in traditional Google search too. The work is never wasted. Your developer can make most of the fixes in a few hours if they know what to look for. glassmark's job is to tell them exactly what to look for, in priority order.

Eighteen months ago, if you'd asked me whether independent hospitality businesses needed to worry about AI-driven discovery, I'd have said "it's coming but you've got time". I wouldn't say that today. The shift has already happened. The question now is whether you get translated into the new language early enough to matter.

Run the scan. See what AI sees.