Your hotel in ChatGPT: What connecting actually looks like

A practical guide to AI distribution for independent hotels — written at the moment before the booking data exists

Nobody has published meaningful booking numbers from AI discovery channels yet. The channel is too new, the sample too small, and any hotel that does have early data has a commercial incentive to be selective about what it shares.

This article is written in full awareness of that gap. What it covers instead is the operational question that comes before the outcome question: what does connecting your hotel to ChatGPT actually involve, what do you control, and what should you be thinking about before you commit to it? If you are running a 60-room independent property and you have been watching the Accor and Marriott announcements wondering where the path is for hotels that cannot build their own ChatGPT app, this is the piece written for you.

What MCP is, in plain terms

The acronym that underpins all of this is MCP — Model Context Protocol. It was developed by Anthropic as an open standard and has since been adopted by OpenAI, Google, Microsoft, and Amazon. That broad adoption matters: it means there is now a single shared language that AI platforms use to request and receive external data.

The simplest way to think about MCP is as a live data bridge. Before it existed, there was no reliable mechanism for a hotel to feed accurate information directly into an AI conversation. The AI would find whatever it could — scraped OTA listings, cached web content, outdated descriptions — and synthesize an answer from those sources. The hotel had no direct input into what was said about it.

MCP changes that by creating a standardized connection between the hotel's data and the AI platform. When a traveler asks ChatGPT for a hotel in your destination, an MCP-enabled integration allows ChatGPT to query your live data directly — current rates, real availability, your own descriptions, your photography — rather than assembling a response from third-party sources. The result is that what ChatGPT says about your property reflects what is actually true today, not what Booking.com wrote about it three years ago.

The protocol is also platform-agnostic. One MCP connection can, in principle, serve any AI platform that supports the standard — ChatGPT today, Gemini and Perplexity as those platforms develop their hotel discovery capabilities.

What Lighthouse built and why it matters for independent hotels

A note before we go further: Lighthouse features prominently in this piece because they were the first to bring this capability to market for hotels of any size. This is not a sponsored article — Hospitality.today doesn't do those — and Lighthouse had no involvement in its production. We cover them here because the product is new, it's the most accessible entry point into this channel right now.

Lighthouse — formerly OTA Insight, the commercial intelligence platform used by a large share of the global hotel industry — launched its Connect AI product in June 2025. The Hotels Network app, the consumer-facing expression of Connect AI inside ChatGPT, launched on March 4, 2026.

The timing matters. ChatGPT's App Directory launched in October 2025, with Booking.com and Expedia among the first partners. From day one, OTA inventory was visible inside ChatGPT — with all of its content accuracy problems and commission implications. Hotels had no direct equivalent. The Lighthouse app is the first open-access path for hotels of any size to establish a direct presence in that same environment.

The structural terms of the product differ materially from OTA distribution and are worth stating clearly:

• The hotel is the merchant of record. The transaction completes on the hotel's own website.
• Lighthouse charges a flat subscription fee. There is no commission on bookings.
• No changes to existing website, booking engine, or property management system are required.
• The app is available globally, with no minimum property size.

The "no integration changes required" claim deserves scrutiny, because in a technology context it is significant. What it means in practice is that Connect AI works as a layer above the hotel's existing systems, pulling content and rate data through Lighthouse's existing commercial intelligence infrastructure. For hotels already using Lighthouse products, this is a relatively light lift. For those that are not, Connect AI ingests the property's data from existing sources — faster than a full integration, but also the point at which content quality becomes commercially significant.

What you control and what you don't

This is the part of the conversation that is least often discussed clearly.

What you control

Your brand narrative is the most significant thing you directly own inside the integration. Connect AI surfaces hotel-provided content — your property descriptions, room type details, amenity information, photography. The traveler asking ChatGPT about your hotel sees your version of your property, not a scraped composite assembled from OTA listings. You can present the recently refurbished restaurant, the accessible room category you added last autumn, the rooftop terrace that never made it onto your Booking.com page. This is the primary differentiator from the default AI response.

Your live rates and availability also feed directly through the integration. When ChatGPT displays your property, it shows current pricing from your own rate structure — not the OTA rate, not a cached figure from last quarter's promotion.

What you don't control

ChatGPT's recommendation logic is not yours to direct. The platform decides which properties to surface based on its own relevance assessment. Connecting through MCP makes your hotel a candidate in that assessment — it does not guarantee placement or ranking. The quality and completeness of the content you feed in will influence the outcome, but the algorithm is not transparent and no hotel or vendor controls it directly.

The booking step also remains outside the AI interface. When a traveler indicates they want to book, they are redirected to your hotel website to complete the reservation. Lighthouse has been explicit that it does not intend to become an in-AI checkout provider. That means the conversion depends on your own booking infrastructure performing well at that moment. A slow-loading engine, a checkout flow that breaks on mobile, a price mismatch between what the traveler saw in ChatGPT and what appears on your site — any of these will cost the booking.

The content question, again

A piece earlier in this series covered the hotel data quality problem in full. The same issue applies here, more acutely: the content you feed into the integration is the content that represents you inside the AI conversation.

Connect AI can only surface what it has access to. If the property descriptions feeding into the system are the same ones sitting on your OTA listing since 2021, the AI response will reflect those descriptions. The integration does not fix content problems — it amplifies whatever content you have. A hotel with rich, current, specific content written in terms travelers actually search for will perform meaningfully better than one that connects the infrastructure and leaves the content unchanged.

The practical checklist before connecting:

• Room type descriptions with enough specificity to answer a conversational query: dimensions, view, bed configuration, included amenities, accessibility details.
• Property-level copy written to answer "what kind of traveler is this for and why?" — not the OTA checkbox format.
• Policies — cancellation, pets, check-in time, parking — current and clearly structured.
• Photography that accurately reflects the property as it exists today.

None of this is new advice. It is the same advice that applies to direct booking conversion optimization generally. In an AI context, it becomes the determining factor in whether your hotel appears at all.

How the connection works in practice

The onboarding process is structured around a demo and setup sequence available at mylighthouse.com/connect-ai. The public documentation does not detail every step of the technical configuration — the specifics vary by property and by which Lighthouse products are already in use. What is clear is that the connection does not require a developer or a system integration project. It is configured at the platform level.

The traveler-side flow once a hotel is connected:

1. A ChatGPT user describes what they are looking for — destination, dates, preferences.

2. ChatGPT queries available apps in its directory, including The Hotels Network app.

3. The app returns hotel-provided data: descriptions, current rates, availability, property details.

4. ChatGPT synthesizes a response.

5. If the traveler selects the property and indicates intent to book, they are directed to the hotel's own booking engine.

That final redirect is the critical step. Everything before it is discovery. The conversion happens on your own platform, which means your direct booking infrastructure — speed, mobile optimization, rate display, trust signals — determines the outcome of every referral the channel generates.

What this does not solve

The vendor communications around AI distribution tend toward enthusiasm rather than qualification. It is worth being direct about the limitations.

Volume is unknown. The channel has been live since March 4, 2026. No hotel has published attributable booking data from AI discovery at any meaningful scale. ChatGPT now has more than 900 million weekly active users globally — but the subset of those users who are actively planning trips, in your destination, for dates you have availability, and willing to follow a redirect to complete a booking on an unfamiliar website, is considerably smaller. What that number looks like in practice is the question the industry will spend the next twelve months answering.

AI discovery is not incremental demand. It is a new surface competing within the existing funnel. The traveler planning a trip to your destination is the same traveler whether they start in ChatGPT or Google. The question is which channel captures them first and what that channel's response says about your property. A hotel without a direct AI presence is not missing an additional audience — it is ceding a growing share of its existing potential audience to whatever the AI finds by default, which right now means OTA listings.

Content quality remains your responsibility. The integration handles the technical connectivity. It does not write your descriptions, update your room types, or refresh your photography. The hotels that see the best results from this channel will be the ones that treat content as ongoing operational infrastructure rather than a one-time setup task.

The case for connecting now anyway

The booking data does not exist yet. The channel is early. The content investment is real. So what is the argument for moving now rather than waiting for the picture to clarify?

It is the same argument that applied to building a mobile-optimized booking engine in 2011, before mobile traffic was large enough to justify it on the numbers alone. The hotels that built for mobile early captured the channel cleanly when it scaled. The ones that waited adapted on the OTAs' timeline, using OTA mobile apps that had already become the default.

The AI travel planning channel is in the same pre-scale phase now. The content you establish, the brand narrative you install in the AI's data sources, and the early presence you build while the channel is still forming all compound over time. A platform that has seen accurate, verified content about your property consistently for a year will represent you differently than one encountering your hotel for the first time after the channel has scaled.

The subscription cost is the relevant variable. If the flat fee is manageable within your technology budget, the case for connecting now is about positioning rather than immediate ROI. If it is not manageable, that is a legitimate reason to wait — but the reason to wait is budget, not skepticism about the direction the channel is heading.

The OTAs have been in ChatGPT since October. Every month without a direct hotel presence in that environment is a month in which the AI's default response to a query about your destination points toward Booking.com.

by Markus Busch, Editor/Publisher Hospitality.today

Read also: The OTA is writing your AI profile. You set it and forgot it.

Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →