What is new
On 29 September 2026, Skift published an interview with Michael Leidinger, senior vice president and chief information officer at Hilton. He spoke on stage at the Destination AI conference in Washington. Skift reports that ChatGPT and other large language models now provide the overall view of the fragmented hotel offering that online travel agencies have provided until now, and therefore put them “under real threat”.
What was said, according to Skift
- A traveller can write a list of requirements in a chatbot and get a working itinerary.
- Hilton sees in large language models and in agentic AI an opportunity for more direct bookings on its own website and in its app.
- Leidinger expects personal AI agents to become channels for which a commission is paid. Skift adds that Meta has said it will, over time, charge for transactions through its agent Muse.
- Agents will check the price up to the check-in date and will cancel and rebook when it drops. As a result, searches will grow faster than reservations and will cost Hilton more in technology.
- People are already giving agents their card details, earlier than Leidinger had expected.
What else Skift wrote
On 2 October it published a second article. In it, Hilton says it has not blocked Muse or any other large agent and accepts reservations from different agents, describing agent reservations as being in “early days”. In the editorial team's tests, Meta's Muse agent makes mistakes on the Hilton, Marriott and Hyatt websites, while another agent, Instinct, reaches the member prices and they match a manual search. An analyst quoted by Skift believes that agents will more often direct guests to online travel agencies, because these hold broad data on availability and prices.
A few figures and where they come from
According to Skift's summary, a survey by McKinsey and Skift Research of 1,000 American travellers shows that half have changed a confirmed stay after seeing a better option. Skift's full text is paywalled; the visible part and the publication's summary have been used here, and the survey itself has not been checked at its original source.
What we do not know
We do not know how many Bulgarian guests ask AI assistants where to stay, or whether such agents will make reservations in Bulgaria. The interview is the opinion of one operator with an interest: Hilton wants more reservations through its own channels.
What is useful for a small property
- Ask for yourself. Put a question such as “where should I stay in [your town]” to ChatGPT or to Google's AI Mode and see whether your place appears and whether its description is correct. Write down the mistakes.
- Keep the essentials at one clear address: rooms with photos, a “from” price if you choose, how to get there, how to contact you and cancellation terms. If an assistant has to answer about your property, it will most likely rely on what is public.
- Do not change your strategy because of a single interview. Over the coming months, keep track of where your enquiries come from.
When ISKA helps
If you do not have your own website, or it does not show the rooms clearly, ISKA Website can show them with photos and an optional “from” price, together with an enquiry form. An enquiry is not a reservation and does not hold a room; a member of your staff confirms it. The website is offered on an individual quote. It does not guarantee a number of enquiries and does not change the platforms' share.
Reading note
Two Skift articles, the paywalled part is not available, the opinion of one operator. Use them as a reason to check your own website, not as a forecast for the market.