BusinessMarkets, money & strategy
OpenAI case study says Radisson's ChatGPT hotel app converts about 1.5 times better than its organic search
Radisson Hotel Group's ChatGPT app, built with Accenture on the Model Context Protocol, searches live rates across 1,000+ hotels. OpenAI reports it converts about 1.5 times better than Radisson's organic search, a vendor figure no independent source has verified.
AT Pulse research deskOctober 8, 2026 · 4 min read · AI-assisted analysisRadisson Hotel Group and Accenture launched the @RadissonHotels app in ChatGPT on 28 July 2026. Users describe what they want in plain language. The app returns matching properties from more than 1,000 hotels in about 100 countries, with live rates, availability, amenities and maps. The booking itself is completed on Radisson's website, not inside the chat. [6] [7] [3]
OpenAI's case study, published 7 October 2026, reports that the app converts bookings at roughly 1.5 times the rate of Radisson's organic search. This is a vendor-reported result. The sample size, time period and definition of conversion are not disclosed, and the comparison leaves out paid channels and online travel agencies. [1] [2]
Technically, the app relies on Accenture's MCP accelerator. It makes Radisson's content, rates and inventory machine-readable and connects them to existing systems for live data. This matches OpenAI's platform pattern, in which an app is an MCP server exposing tools that must pass a review. Tool schemas, ranking logic and latency are not published. [7] [8] [4]
Radisson joined a crowded field. Booking.com was a ChatGPT launch partner in October 2025, and Accor launched a ChatGPT app with loyalty-member rates in January 2026. Wyndham and IHG followed, and Marriott launched its own conversational search tool. Booking inside the chat, managing reservations and loyalty recognition remain planned features for Radisson. [3] [5] [4] [7]
Accor and Radisson run ChatGPT apps that send bookings to their own websites, and IHG, Wyndham and Marriott have launched conversational tools too.
Read the full assessment
Implication: chat assistants could become a direct-booking channel competing with online travel agencies, though performance data remain vendor-reported.
Sources are collected; writing the dossier now.
Executive brief
OpenAI says Radisson's ChatGPT app converts bookings at about 1.5 times the rate of Radisson's own organic search. That figure comes from OpenAI's case study, as reported by tbreak, and no independent source has checked it. Radisson Hotel Group and Accenture launched the app, called @RadissonHotels, on 28 July 2026. It searches more than 1,000 hotels in about 100 countries using live rates and availability. Travellers still finish the booking on Radisson's website. OpenAI's write-up on 7 October 2026 presents it as a reference case for agentic commerce, a market where Accor, IHG, Wyndham and Marriott are also building (Voyages d'Affaires).
What changed and event timeline
OpenAI opens ChatGPT to third-party apps
At DevDay, OpenAI previewed the Apps SDK, which is built on the Model Context Protocol (MCP). Booking.com was one of the launch partners.
Accor ships first
The ALL Accor app went live in ChatGPT in more than 20 languages, showing loyalty-member rates and sending users to Accor's own site to book.
Radisson launches @RadissonHotels
The app was built with Accenture and covers 1,000+ hotels with live inventory, rates, amenities and maps.
More detail
OpenAI publishes its case study
OpenAI's customer story describes a "ChatGPT plugin" for finding, comparing and booking hotels, and reports early conversion results.
More detail
Capabilities and access
- How to use it: type "@RadissonHotels" in a ChatGPT conversation and ask in plain language, for example a Paris hotel with a spa near the Eiffel Tower (Voyages d'Affaires).
- What it returns: matching properties with live rates and availability, amenities, location context and an interactive map (eHotelier).
- Booking: the final booking happens on Radisson's website, not inside the chat.
Read the full section
- How to use it: type "@RadissonHotels" in a ChatGPT conversation and ask in plain language, for example a Paris hotel with a spa near the Eiffel Tower (Voyages d'Affaires).
- What it returns: matching properties with live rates and availability, amenities, location context and an interactive map (eHotelier).
- Booking: the final booking happens on Radisson's website, not inside the chat.
- Model: no model or version is named in the reviewed sources.
- Regions: coverage by region is unclear. tbreak notes that hotels in the Gulf/UAE are not specifically confirmed.
Technical analysis for researchers and developers
- Architecture: the app uses Accenture's MCP accelerator, part of its "AI Merchant Center."
- Platform pattern: in OpenAI's current developer docs, a plugin is an MCP server that exposes tools.
- Not published: the tool schemas, the way searches are ranked, response times and any evaluation method. No one can reproduce the conversion result.
Read the full section
- Architecture: the app uses Accenture's MCP accelerator, part of its "AI Merchant Center." It formats Radisson's content, rates and inventory so an AI can read them, and connects to Radisson's existing systems for live data (eHotelier).
- Platform pattern: in OpenAI's current developer docs, a plugin is an MCP server that exposes tools. Selected tools can return UI elements shown inside ChatGPT. Public plugins must pass a review of the remote MCP server and the directory guidelines (OpenAI developer docs).
- Not published: the tool schemas, the way searches are ranked, response times and any evaluation method. No one can reproduce the conversion result.
Claims and evidence
- 1,000+ hotels, ~100 countries, live rates — Vendor-reported ()
- ~1.5× conversion vs. organic search
- 87% of travellers willing to let AI handle hotel search
Read the full section
| Claim | Status |
| 1,000+ hotels, ~100 countries, live rates | Vendor-reported (Nasdaq) |
| ~1.5× conversion vs. organic search | Vendor-reported (OpenAI). The sample size, time period and definition of "conversion" are not disclosed (tbreak) |
| 87% of travellers willing to let AI handle hotel search | Survey cited in the press release, with no method stated (eHotelier) |
| App is live and booking happens on Radisson's site | Consistent across several outlets, but they all draw on company materials |
No independent check of performance was found.
Context and prior work
- Radisson came in after other hotel groups.
- Booking.com was a ChatGPT launch partner in October 2025 (Constellation Research).
- Accor's app shows loyalty-member pricing (AltexSoft). Radisson lists loyalty recognition as a future feature.
Read the full section
- Radisson came in after other hotel groups. Accor launched in January 2026, Wyndham and IHG followed, and Marriott launched its own "Ask Bonvoy" conversational search in June (Voyages d'Affaires).
- Online travel agencies got there first. Booking.com was a ChatGPT launch partner in October 2025 (Constellation Research).
- Accor's app shows loyalty-member pricing (AltexSoft). Radisson lists loyalty recognition as a future feature.
Limitations, safety and contested findings
- Weak baseline: the 1.5× figure is compared only with organic search, not with paid channels or online travel agencies.
- Unfinished features: booking inside the chat, managing or changing reservations, loyalty recognition and AI concierge services are all planned, not live (eHotelier).
- Unclear naming: OpenAI calls it a "plugin" while the press release calls it an "app" (OpenAI).
Read the full section
- Weak baseline: the 1.5× figure is compared only with organic search, not with paid channels or online travel agencies. It also comes from the platform vendor (tbreak).
- Unfinished features: booking inside the chat, managing or changing reservations, loyalty recognition and AI concierge services are all planned, not live (eHotelier).
- Unclear naming: OpenAI calls it a "plugin" while the press release calls it an "app" (OpenAI).
Business and practitioner implications
- Distribution: hotel chains see ChatGPT apps as a way to win direct bookings without depending on third-party platforms (Voyages d'Affaires).
- Build effort: most of the work is preparing data, meaning clean, machine-readable rates, inventory and content behind an MCP server.
- Measurement: before relying on vendor figures, teams should define conversion and attribution for chat-referred traffic themselves and compare it against paid channels.
Read the full section
- Distribution: hotel chains see ChatGPT apps as a way to win direct bookings without depending on third-party platforms (Voyages d'Affaires). Online travel agencies are competing in the same chat interface.
- Build effort: most of the work is preparing data, meaning clean, machine-readable rates, inventory and content behind an MCP server. Systems integrators such as Accenture are packaging this as accelerators.
- Measurement: before relying on vendor figures, teams should define conversion and attribution for chat-referred traffic themselves and compare it against paid channels.