On October 7, OpenAI published early results from Radisson Hotel Group’s ChatGPT app — the first concrete figures anyone has put on what a branded app inside an assistant actually produces. Across July and August, traffic arriving through the app converted to bookings at roughly 1.5 times the rate of Radisson’s own organic search. Velit Dundar, the group’s VP of global ecommerce, called the results “encouraging.”
The build is the other number. Radisson and Accenture shipped the thing in six weeks.
What they actually built
The app went live on July 28 and covers more than 1,000 properties in over 100 countries. A traveller describes a trip and gets live inventory, rates, amenities and map-based results inside the conversation, then clicks through to Radisson’s own site to complete the reservation. Underneath sits an MCP server Accenture built against Radisson’s content, inventory and rates — the same plumbing now feeding its ChatGPT ads.
Two details matter more than the headline. First, invocation changed: at launch you had to type @RadissonHotels, and after a ChatGPT update the app now surfaces on its own when it is relevant and installed. Second, the honest caveat, which OpenAI’s own write-up includes: 54% of recorded checkout and booking events were attributed to ad views. Radisson is buying sponsored placement in ChatGPT alongside the app, so 1.5x reflects both, not free organic lift.
Why an events team should care about a hotel story
Because the delegate journey is the same journey. Someone deciding whether to attend your conference is planning travel, comparing dates, checking what a pass includes and where to stay — increasingly in one conversation with an assistant rather than across eight tabs. Radisson’s result says that conversation converts better than search when the brand has published machine-readable facts into it.
It also shows where the traffic lands. The assistant handled discovery and then handed a half-decided person to Radisson’s own channel to finish. For an organiser, that means more people hitting your registration page already warm and asking the specific questions — group rates, cancellation terms, what the pass covers, accessibility, invitation letters — at 11pm on a Sunday, six weeks out. That volume spike around registration and event week is the oldest problem in running events and conferences, and assistants make it arrive faster and later in the day.
What to do before your next registration opens
Start with the data, not the agent. Put dates, venue, pricing tiers, what each pass includes, cancellation terms, the hotel block and accessibility details on crawlable pages in plain text. A PDF brochure or an image of your agenda is invisible to the thing doing the recommending.
Then pull last cycle’s inbox and phone log and count the five questions that repeated most. Those are your script. Radisson needed six weeks and an MCP build to answer them in a chat window; you need your own site, phone line and messaging channels to answer them consistently, which is a much shorter job. Something that answers across web, phone and messaging at once, captures the sponsor and group enquiries it cannot close, and routes the rest to a named human, covers most of it without adding registration-desk headcount.
Last, measure one number before you change anything: how many enquiries arrive outside office hours, and how many of those you never reply to. Radisson’s 1.5x is a lift on traffic it already had. The enquiry nobody answers at 2am converts at zero, whatever your funnel looks like.
Working in events & conferences? See how TalkRev works for Events & Conferences.
The TalkRev blog
One short email per new post
No ads. Unsubscribe anytime.
