Evidence at a glance
Moving Hotel Choice Ahead of the Website
Radisson Hotel Group and Accenture Song used OpenAI technology to build the Radisson Hotels ChatGPT app, where travelers can search for and compare hotels based on destination, amenities, and trip needs. It is aimed not at people already on a hotel website filling out a booking form, but at travelers still planning a trip and narrowing their options.
The shift matters not because guests can now complete a booking in chat, but because the brand and its inventory appear earlier in the decision process. Users still leave ChatGPT to book on Radisson’s website, so the competition turns on two questions: whether a hotel can surface as a relevant option in a conversation, and whether that discovery can lead to the company’s own direct channel.
Behind the Conversation Is an Inventory Connection
After a traveler describes a need in natural language, the app returns hotel options and information such as maps, location, amenities, prices, and availability. Accenture says the app connects to Radisson’s existing systems for live hotel information. Before that connection, its AI Merchant Center MCP accelerator was used to structure, validate, and optimize hotel content, inventory, pricing, and booking signals.
The challenge is therefore not only to interpret a request such as “somewhere central with a pool,” but to map it to hotel data that is reliable and current. The public materials do not disclose the API fields, model-call sequence, or deployment architecture, so those details cannot be inferred. What they do describe makes clear that organizing and connecting supply data is the foundation for turning conversational intent into options people can compare.
Discovery Happens in Chat; Checkout Stays Direct
Radisson has not moved the booking transaction into ChatGPT. When travelers are ready to book, they are sent to Radisson’s website. The app handles discovery and comparison, while the website remains responsible for checkout and the direct customer relationship. For the hotel group, this extends its entry point: it can meet prospective guests before they visit its site while trying to keep the eventual booking in its own channel.
That split creates a new dependency. Information shown in the conversation has to line up with prices, availability, and the booking flow on the website; inconsistencies between the two interfaces can undermine the recommendation. Accenture says the project took six weeks to build and launch, but the public materials give neither staffing nor cost figures. The timeline shows that the experience could be shipped quickly, not that ongoing data maintenance and cross-channel operations are equally easy.
The Conversion Figures Are Promising, but Bounded
OpenAI reports that, in July and August 2026, the app’s visit-to-booking conversion was about 1.5 times Radisson’s organic-search conversion. That is an encouraging early signal, but the case study provides no sample size, baseline conversion rate, or statistical significance. It also does not say whether the two traffic sources had comparable user intent. The figure suggests that this incoming traffic performed well; on its own, it cannot show that ChatGPT generated bookings beyond those that would otherwise have come through organic search.
Another reported figure is that 54% of recorded checkout and booking events were attributed to ad views. View-through attribution does not mean users clicked an ad before booking, much less that the ad created those bookings. Coverage figures also use different scopes: Accenture says the app can search more than 1,000 hotels in over 100 countries, while OpenAI describes Radisson as having more than 1,640 operating and developing hotels across EMEA and APAC. These are not equivalent counts.
Teams Need to Test More Than Whether People Book
For technology and digital-business leaders, the first priority is not a more elaborate chat experience. It is making sure that the hotel data exposed to the app is complete, current, and consistent with the website’s booking flow. Radisson has worked with Accenture on its direct digital channel for seven years; the ChatGPT app is best understood as a new discovery layer in front of that channel, not a replacement for its booking infrastructure. Bringing the branded app and sponsored advertising together also makes this a distribution and acquisition question, not just a product-design one.
Evaluation should separate performance from incrementality. Alongside conversion, teams should track the full path from conversational recommendation to website visit and booking, then use a control or other incremental test to establish whether the new entry point changes total bookings rather than merely relabeling their source. They should also examine how prices and availability stay synchronized, how inaccurate information is handled, and how ad-view attribution windows are defined. The public materials do not answer those operational questions. Until they are tested, the sound conclusion is to treat ChatGPT as a promising upstream route into direct booking, not as a proven new growth engine.