AI sends hotel websites less than one visit in a hundred. If that number is making you relax, you are reading the wrong number.
The figure comes from a sharp piece by Markus Busch at Hospitality.today, published on 7 October 2026. His conclusion is sensible: the AI guest is too small to rebuild anything around this winter, and growing too fast to ignore. I agree with half of that. The half I disagree with is the bit most hotel teams will hear, which is “too small”.
The small number
Here is what the data says. Similarweb figures, published by Bain & Company on 30 September, show that the share of visits AI platforms sent to hotel and short-term rental websites quadrupled between July 2025 and July 2026. Even so, every travel category Bain looked at, including OTAs and cruise lines, stayed under 1% of total traffic.
Booking Holdings tells a similar story. On its August earnings call, the company said traffic from AI assistants, paid and unpaid together, was significantly below 1% of its room nights, and that the share had not really moved for quarters.
So far, so reassuring. Under 1% is a rounding error in most board reports.
The numbers that should worry you
Now look at the people, not the clicks. Phocuswright found that 56% of US leisure travellers used AI for at least one trip in the past twelve months, up from 33% in the first half of 2025. Bain’s consumer research says 49% of US travellers who have used AI for travel now rely on it most or all of the time. In the UK, France and Germany, the figure is 41%.
Put those side by side. More than half of travellers are planning with AI. Less than 1% of visits come from it. That gap is not proof that AI does not matter. It is proof that most of the work happens where your analytics cannot see it.
A traveller asks ChatGPT for a quiet seaside hotel with a good restaurant. She gets three names. Two days later she types one of them into Google, or opens an OTA app, and books. In your reports, that looks like brand search or direct traffic. The AI did the shortlisting. Google gets the credit.
And then there is conversion. Adobe’s figures, also covered by Hospitality.today, show that in July 2025 visitors arriving from AI sources converted 47% less often than everyone else on US travel sites. By July 2026, the gap was just 1%. AI traffic to those sites was also up 119% year on year.
A year ago, the chatbot visitor browsed and left. Now she books like everyone else. A small channel with average conversion and fast growth is not a channel you ignore. It is a channel you get ready for before it becomes expensive.
We have seen this film before
Travel has a habit of waiting until a channel is big before taking it seriously, and then paying someone else to get in. Hotels did it with OTAs. Kayak is now living it with Google, as we covered when the metasearch bosses admitted free traffic is drying up.
The AI version is already taking shape. Google has written the first draft rules for AI hotel booking, and independent hotels were not at the table. Brands like Princess Cruises are already feeding their data straight into ChatGPT and Claude. By the time AI referrals hit 5%, the shortlists will be set. If the assistant cannot read your rates and rooms clearly, it will happily send the booking to an OTA instead.
Be honest about the data, too
To be fair to the sceptics, every one of these numbers has a catch. Similarweb tracked only three to five sites per category, so the Bain figure is about big brands, not a forty-room independent. Bain does not say when its survey ran or how many people it asked. Booking measures its own traffic and has every reason to sound calm. Adobe sells tools that track AI traffic.
That is exactly why you should not borrow anyone’s number. Measure your own.
What I would do this month
- Find the AI guest in your analytics. Filter the last twelve months for referrals from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com. Count the visits, compare conversion with Google organic, and note the landing pages.
- Watch brand search alongside it. If AI referrals are flat but branded searches and direct visits are climbing, the AI may be doing the shortlisting while Google takes the credit.
- Ask the assistants about yourself. Check how ChatGPT, Gemini and Copilot describe your property. Are the price, location and facilities right? If not, fix the sources they lean on.
- Make the landing page do its job. AI visitors arrive having already read about you. They want to confirm the price, the room and the policy, fast. Give them that.
None of this is expensive. Most of it is the same groundwork I set out in how a travel CMO should prioritise SEO budget. The difference is that you start tracking it now, while it is cheap to learn.
The debate
So here is my challenge to every hotel marketer who has seen the “under 1%” headline and moved on. Is AI really a tiny channel, or is it a huge influence hiding inside your brand search numbers?
I think it is the second. And I think the hotels that wait for the referral number to look big will find the shortlists were written without them.
Follow the newsletter: Dan Taylor on LinkedIn