Last week, fifty hotel marketers walked into 12th Knot at Sea Containers London, with the Thames doing its best to distract everyone.
After some breakfast and a bit of networking, our Managing Partner, David Gardner opened with a data-led look at where websites stand today. A short break later, Sam Weston, Head of AI & Marketing at 80 DAYS, took the stage to moderate a panel with Anna Wolfram, Head of Marketing at Grantley Hall, Olivia Smith, Digital Marketing Lead at The Savoy and Brian Gore, Group Director of Marketing and Communications at Arora Group.
The question running through both was a somewhat provocative one: do hotels still need a website?
Coming from us, that takes a certain nerve. We build websites for a living. But it's a fair question, and the answer turned out to be far more interesting than a simple yes.
A dilemma over coffee
Before the panel got going, we put a choice to the room. Would you rather have a website that's brilliant for humans and mediocre for AI, or brilliant for AI and mediocre for humans?
It wasn't close. Of the 37 who voted, 29 chose humans, 8 chose AI.
Good instinct. But hold that thought, because the same room told us something rather different a few minutes later.
We've been measuring the wrong thing
David opened with an industry-wide confession. "For 25 years, we’ve been confusing how important a website is with how much traffic goes into it," he said. Traffic climbs 10% and everyone's delighted. It slips 1% and you're in a meeting with the GM explaining yourself.
That habit is about to become a problem, because the traffic is genuinely changing shape. Look at the London four and five-star hotels in our own Hotel Benchmark platform: website visits down 16% year on year, website bookings up 2%. Fewer people arriving. The same number booking.
David's summary of it fitted succinctly on one slide. Fewer visits, higher stakes.
Why? Because the homework is being done elsewhere. The old journey ran search, click, research, compare, shortlist, book, and your website kicked in early. The new one runs ask, shortlist, verify, book, and your website turns up near the end, when someone is close to spending real money.
Which means the visitor has changed too. They arrive with dates, a budget and an occasion in mind. Adobe's numbers back this up: people referred by AI spend around 70% longer on travel sites, with engagement up 21% and bounce rates down 41%.
As David put it: "AI can build the shortlist. Your website still has to build the conviction."
Positioning, before anything technical
There was a moment when David put a sentence on screen that could belong to almost any hotel on earth. A luxury hotel offering exceptional service, unforgettable experiences and a perfect location.
"If a person can't tell which hotel that is, AI hasn't got a hope," he said. "It's intelligent, but it's not psychic."
He pointed to research in which 2,700 questions were put to six AI platforms about twenty-five London luxury hotels. Nearly 60% of recommendations went to just five of them, the five that were crystal clear about who they were. One 26-bedroom independent was outperforming major chains. That simply wouldn't have happened before.
His conclusion is perhaps the most quotable line of the morning: "AI doesn't create a positioning problem. It just exposes one that you may have."
The room's confidence gap
Then we asked the second question. How confident are you that AI describes your hotel accurately right now?
Of the room, 12 said it gets them right. 23 admitted they were nervous. 2 had no idea.
So a room that overwhelmingly wants to serve humans first is also, by a clear majority, unsure whether the machines are describing them properly. It's just an honest gap, and the panel filled it with specific examples.
Anna Wolfram of Grantley Hall found that AI had quietly promoted their country pub into a sixth restaurant. They talk about having five restaurants, three bars and a village pub, but AI had folded the pub into the restaurant count and dropped it from the description entirely. For Anna, that was a clear sign they weren't differentiating the pub well enough themselves.
At the start of the year, Olivia Smith at The Savoy had been fielding guest complaints about afternoon tea prices that were five years out of date. Brian Gore at Arora Group found AI listing a rebranded restaurant twice, the old name alongside the new one, for what is physically the same room.
Anna's advice was some of the most actionable of the session, and she's already automated it. She once tracked sixty-two listings and profiles by hand, cross-checking them every six months. "Now I ask AI to flag anything inconsistent and it's done in five seconds. It even alerts me weekly if something new crops up."
Inspiration and information can share a house
The obvious worry is that feeding facts to machines will sand the soul off a beautiful site. The panel didn't buy it. The trick is layering the two together.
Anna explained how Grantley Hall does it in practice: the top of the page sets the scene with video and lovely adjectives, while the FAQs and terms sit further down, doing the less glamorous work of accuracy
Brian agreed. The FAQ page, largely abandoned in recent years, is back doing serious work, a human might never read it, but AI will.
Olivia put it plainly about FAQs: unglamorous, and not something to dress up. They just need to be practical and answer the question. Her broader point, on where AI can't compete, stuck with the room: it might see two historic London hotels as interchangeable, but a website is where a hotel gets to be different. That difference doesn't come from another photo of a bedroom, it comes from the small, specific things guests don't expect.
She also flagged the brutal new maths of discovery: there's no page two in AI. With only around five slots up for grabs, missing out means missing the customer entirely.
The agents are coming. That's the opportunity.
IDC reckons 30% of travel bookings will be made by AI agents by 2030. We asked the room how they felt about AI agents: 18 called it a threat, 9 saw it as an opportunity, 6 stayed sceptical and 4 admitted they were just exhausted. Fair enough on that last one, the pace of change in AI is exhausting.
Our own testing gives the nervousness some grounding. In roughly a third of cases, an AI agent hit enough friction on a hotel site, cookie banners, interstitials, fiddly booking steps, that it gave up and went to an OTA instead. We're now exploring a new web standard called WebMCP to try and fix exactly that.
But Brian made the case for optimism better than we could. "We handed ourselves over to the OTAs once. This might be our chance to take some of that back, if we get it right."
Anna has taken the boldest route available and left the OTAs entirely. Four months in: no drop in bookings, and no commission either.
And Brian's practical advice? "If AI and GEO aren't in your daily digital conversation, put them there. Don't wait six months. Now is the time."
So, five years from now?
Finally, we asked the panel to get their crystal balls out: what does a hotel website look like, five years from now? Brian went with AI-powered and conversational, guiding each guest in real time. Anna predicted many versions of one website, shaped around who's looking. Olivia chose one word: aspirational, letting someone genuinely picture themselves there.
Sam tied it together in summary: one website, two audiences. Inspiration for the humans, and underneath it all, everything a machine needs to know.
The website isn't going anywhere. It's just being asked to do a harder, more valuable job. AI can recommend a hotel. Your website still has to make you want to stay there.
Thanks to Anna, Olivia and Brian for such a generous, candid session and to everyone who joined us.
So where do you start? David’s session gave five places to start this week:
Audit your positioning. If it's not clear, nothing else you spend on the website will make up for it.
Audit your proof. Room sizes, directions, the hard facts AI needs to work with. Make sure they're easy to find.
Test the decision journey. Pick two close competitors and see how you compare when someone's actually trying to choose.
Check your AI visibility. Go and look at what AI is currently saying about you, and whether it's right.
Reset your measurement. The old questions (traffic, clicks) aren't enough on their own. Start asking whether your visibility is creating qualified demand, and whether that's converting into direct bookings.
Positioning, proof, decision journeys, AI visibility, measurement, that's the work we do with hotels every day. Get in touch and we’ll show you where to start.