What actually launched, and when
Three things happened inside two weeks this summer. None of them is a thing a hotel group can go out and buy, but together they show where the choosing is moving.
| Who | When | What it does | Where the booking completes |
|---|---|---|---|
| American Express Global Business Travel | 27 July 2026 | An Egencia connector for the Claude assistant that lets employees and company AI agents "search, book and manage policy-compliant flights and hotels without opening a booking tool" | Inside the assistant. The company calls it one of the first in managed business travel to complete a booking end to end rather than just hand back information |
| IHG Hotels and Resorts | 28 July 2026 | Conversational search across more than 7,000 hotels, in test in the United States on its own website and app. You describe the trip, it returns hotels with live availability and prices | On IHG's own channels |
| Radisson Hotel Group with Accenture | Reported 6 August 2026 | A Radisson app inside ChatGPT covering more than 1,000 properties in over 100 countries, with live rooms and rates | Not in ChatGPT. The guest is sent to Radisson's own website to finish |
That last row is the one worth sitting with. Even the hotel groups building this are, for now, using the assistant to get chosen and their own channels to get paid.
The shortlist is where it is decided
If you run one hotel, or eight, the change that reaches you is quieter than the headlines. Your guest arrives having already been handed a shortlist, and the shortlist is short.
Source: eHotelier, "AI is already choosing which hotels get booked", July 2026. A panel description of the shift, not a survey figure.
The second thing those panellists said matters more, because you can do something about it today. An assistant does not read your writing the way a guest does. It reads your facts. Your room types, what is actually in the room, whether you take dogs, what time you let people in, what the parking really costs. As one of them put it, it does not understand brand voice, it understands structured truth.
So if those facts disagree between your own website, your listings on the booking sites and your Google entry, you are a harder hotel to recommend with confidence. It is unglamorous work, and it is the part that decides whether you make the four. Making the same true thing appear everywhere, and keeping it current.
Being straight with you
Sorino does not put your hotel in that shortlist. No one can honestly promise that. There is no slot to buy and no switch to flip. What we can tell you about is what happens to the guests who come out of it, because that part lands on your phone and in your inbox.
The four that made the list all get asked the same question
Here is what a shortlist of four does to your phone.
The guest has been given your name alongside three others, and they are loyal to none of you yet. They want one or two things settled before they commit: can you do a family room on those dates, is the restaurant open on the Sunday, can they get in early. So they ask. By phone if they are older, by web chat or a message if they are not, and often late in the evening, because that is when people actually plan trips.
Four hotels get that question. The one that answers it properly and quickly, with a real answer rather than "we will come back to you", is usually the one that gets the booking. That has always been true. What is different is that there are now three competitors instead of thirty-nine, and the guest is comparing you side by side.
This is the part that lands on your desk, and it is the part an always-on agent is for. It picks up day and night, answers from what is actually free, holds the room, and puts the enquiry in one place whether it came by phone, email, web chat or social. And it never guesses: if it is not completely sure about an allergy, a dog, an access need or anything else that would embarrass you on arrival, a real person steps in.
Five things to put on someone's desk this month
Each of these is a job you can hand to a named person, across every property.
- Audit your ten guest facts, everywhere they appear. Room types and how many each sleeps, pet policy, parking, check-in and check-out times, breakfast, accessibility, restaurant hours, cancellation terms, what is included. Check them across your own website, your booking engine, your Google entry and every booking site listing. Write the true version once and fix the rest.
- Give each property one owner for those facts. In a group, this is where it falls down. A rate or a policy changes at one hotel and four listings quietly go stale. One named person per property, checked quarterly.
- List the ten questions you actually get asked, and see who answers them. Then check when, and how fast. Most groups have never looked at this property by property.
- Test a Sunday evening enquiry yourself. Nine in the evening, from a mobile, once by phone and once by web chat. Time the reply. That is the test the shortlist applies, and it is the only one that matters.
- Decide what may be answered automatically and what must always reach a person. Rooms and rates from live availability, yes. Allergies, access needs and anything about a guest's health, no. Write that line down before you automate anything, not after.
Doing the first two is worth it whichever way the technology goes, because consistent facts also make you easier to book the ordinary way.
Frequently asked
Will guests actually book through an AI assistant? In business travel, some already can. For leisure, the pattern so far is that the assistant helps the guest choose and then sends them to a website or a phone to finish. Radisson's ChatGPT app works exactly that way.
Does this replace the booking sites? Not yet. It adds another place where the choosing happens before anyone reaches your website.
Can I pay to be in the shortlist? There is no advertising slot to buy today. What the people working on it recommend is unglamorous: consistent, accurate, structured facts about your hotel wherever it appears.
We are a small group without a technology team. Is this out of reach? The first four things on the list above are a morning's work with a spreadsheet and someone who knows the hotels. None of it needs a developer.
This is the hotels edition. There is a wider set of reads across hotels, restaurants and bars. Read more on the Sorino blog →
The short version
- Assistants are starting to search and book hotels: Amex GBT's Egencia connector for Claude can complete a business booking end to end, and IHG and Radisson have both launched conversational search.
- Even the groups building it use the assistant to get chosen and their own channels to get paid.
- The guest's shortlist is compressing from around forty hotels to about four.
- Assistants read facts, not marketing copy, so your room types, policies and prices need to agree everywhere they appear.
- Nobody can sell you a place in the shortlist. What you control is how fast and how well you answer the guests who come out of it.
Sources
- Amex GBT launches an Egencia AI connector for Claude, letting employees and enterprise AI agents "search, book and manage policy-compliant flights and hotels without opening a booking tool", 27 July 2026. Travelling for Business. travellingforbusiness.co.uk
- IHG Hotels and Resorts launches AI conversational search across more than 7,000 hotels, 28 July 2026. IHG plc. ihgplc.com
- Radisson Hotel Group launches a ChatGPT hotel app with Accenture, covering more than 1,000 properties, with bookings completed on Radisson's own website, reported 6 August 2026. ChannelLife UK. channellife.co.uk
- "AI is already choosing which hotels get booked: 5 takeaways from top industry leaders", including the shortlist compressing from around forty hotels to four, and the point that AI "understands structured truth", 23 July 2026. eHotelier. insights.ehotelier.com