ChatGPT hotel booking: what a chain's ChatGPT plugin changes for an independent hotel

On 7 October 2026 OpenAI published a customer story titled "Radisson Hotel Group brings hotel discovery into ChatGPT."

We read it in a browser on 12 October 2026, because a plain request to the page came back 403.

The headline number is this: OpenAI's Radisson case study says early visit-to-booking conversion from the plugin ran at "approximately 1.5 times" Radisson's organic search for July to August 2026.

That is a vendor's case study about a paying partner. Treat it as a claim, not a benchmark.

The same week, Google autocomplete for "chatgpt hotel" finished with booking, search and finder. So owners are asking the obvious question.

If a chain has its own window inside ChatGPT, where does a twelve-room property stand?

Here is the complication. The answer to "can ChatGPT book a hotel" is still mostly no, and the Radisson plugin does not change that. What it changes is who gets shown before the guest leaves the chat.

Step 1: read what the plugin does

OpenAI's page calls it a plugin, not an app. The link on the page points to a ChatGPT plugin listing whose address carries an Apps SDK identifier.

The case study says the plugin "lets users see Radisson Hotel suggestions on a map, compare pricing, and view hotel amenities in detail directly in ChatGPT."

Then the guest continues "to the Radisson site to complete their booking."

So the booking is not in the chat. The shortlist is.

Accenture built it, and the case study says it took six weeks. It also says Radisson runs sponsored ads in ChatGPT alongside the plugin, and that more than half of recorded checkout and booking events from those campaigns were attributed through view-through measurement. That is ad-view attribution, not clicks. Read it that way.

Step 2: notice the line that matters

One paragraph in the case study deserves more attention than the conversion figure.

It says a recent ChatGPT update means users no longer need to type @RadissonHotels. If they have the plugin installed, ChatGPT "will automatically show it whenever relevant."

That is the shift. A user who installed a chain's plugin can get that chain's map, prices and amenities without asking for the chain by name.

Note the condition: installed. OpenAI's help page Connected apps in ChatGPT, read on 12 October 2026, says you set one up from Settings, then Plugins, and can call it with @ or the plus menu. It also says app availability depends on the app, the plan, the region, the workspace and the interface.

So the plugin wins inside a subset of conversations. Everyone else still gets web search.

Step 3: understand how a chain plugin gets picked

OpenAI tells developers how ChatGPT chooses a tool.

Its developer guide Optimize Metadata, read on 12 October 2026, says ChatGPT and Codex "decide when to call your tool based on the metadata you provide."

It tells builders to test three prompt sets: direct prompts that name the product, indirect prompts that describe the outcome, and negative prompts where another tool should answer.

Read that from the guest's side. "Somewhere quiet near the old town with a desk" is an indirect prompt. A chain that tunes its metadata well can catch it.

An independent property has no metadata in that layer. It has a website.

Step 4: where the independent gets found instead

Without a plugin, a hotel reaches a ChatGPT answer the same way a dentist does: through search.

OpenAI's help page Searching the web with ChatGPT, read on 12 October 2026, says ChatGPT "typically rewrites your query into one or more targeted queries" and may share the user's general location with outside search providers.

It ranks results "using multiple factors intended to help users find relevant, reliable information." And: "Placement is not guaranteed."

We covered what that page means for any business in is ChatGPT a search engine, so we will not repeat the whole pipeline here.

The OTAs sit in this same layer. Their listing pages are web pages, and ChatGPT search can retrieve them. People also type "chatgpt booking.com" into Google. We did not find an OpenAI or Booking.com page confirming a Booking.com plugin in ChatGPT when we checked on 12 October 2026, so we are not telling you one exists.

Step 5: stop treating this as a plugin race

A plugin is the wrong first move.

That will sound odd from a GEO shop. But look at what the Radisson plugin needed: a partner network relationship, an agency, an MCP server and APIs, and a guest who installs it.

An independent hotel can match none of that, and does not need to. The guest who has not installed any chain plugin, which on OpenAI's own description is the default state, still gets an answer built from the open web.

NovaTechRay's view is that the open-web answer is where an independent can still win, because it is decided by facts on pages rather than by who shipped software. We could be wrong about how long that holds. OpenAI's case study says Radisson plans more functionality inside ChatGPT, and we cannot see the roadmap.

Step 6: fix the three things you control

Crawler access. The search help page says that to be eligible, a site must allow OAI-SearchBot and confirm its host or CDN allows traffic from OpenAI's published searchbot IP addresses. If your booking engine vendor or CDN blocks unfamiliar bots, you are invisible before ranking starts. The robots.txt lines are in our GPTBot, OAI-SearchBot and ChatGPT-User breakdown.

Rates on your own page. The plugin's selling point is "compare pricing." Your counter is a rate range in plain HTML, seasonally caveated, outside the booking widget. A crawler cannot read inside most widgets, and a guest's question almost always contains a budget. We argued the same in hiding your prices keeps you out of AI answers.

Structured facts. The plugin shows amenities "in detail." Match it in text: room sizes, bed set-ups, step-free access, breakfast hours, desk or no desk, measured wifi. Then mark it up with Hotel and HotelRoom schema. The full attribute list is in AI search visibility for boutique hotels.

None of this is new. What is new is that the chain next to you now gets a map with prices for some users, so a vague page loses by a wider margin. A page comparing you with that chain is the next thing to write.

Step 7: check what you can check this week

Open ChatGPT logged out, so no plugins are installed. Ask the three questions a guest would ask about your area, with a budget and a constraint in each.

Note whether you appear, and which page was cited: yours, an OTA, or a guide. GA4 will undercount any clicks those citations send, for the reasons in our referral corrections.

Then check your server logs for OAI-SearchBot requests. If there are none, start with your CDN's bot settings before you touch any copy.

NovaTechRay starts a hotel engagement with this check, because a firewall rule can hide a page that no paragraph will fix.

One caveat on all of it: ChatGPT changes fast, and a logged-out test is one sample. Repeat it monthly.

Our group blog keeps a written method for reading announcements like OpenAI's case study, including the six checks every VALORAE post runs on a platform claim.

The short version

• Read the Radisson case study as a vendor claim: the chat shortlists, the chain's own site books.
• Remember the plugin surfaces automatically only for users who installed it.
• Allow OAI-SearchBot and OpenAI's published IP addresses at your CDN and booking engine.
• Put a plain-text rate range on your own page, outside the widget.
• Write room-level facts in text and mark them up with Hotel and HotelRoom schema.
• Test logged out every month and log which page ChatGPT cites.

This is the work NovaTechRay does at novatechray.com.