On Sunday, May 24, 2026, Memorial Day weekend, OpenAI sent an email to ChatGPT Plus and Pro subscribers announcing that ChatGPT could now use connected Gmail, past chats, saved memories, and personal files to generate more personalized responses. The travel use case was front and center. The example in the announcement: “Based on what you know about me, where should I travel next?”
A Sunday afternoon on a holiday weekend is not when companies announce things they want scrutinized. It is when they announce things they want absorbed quietly.
ChatGPT answered. It referenced travel history and personal style. It named a destination.
For independent luxury hotels, the issue is not whether ChatGPT can recommend travel. The issue is who is already connected to the recommendation layer when that answer is formed. Expedia is already in the room.
This Is Not Speculation. It Is a Published Commercial Reality.
Before asking what the personalization update means for hotel demand, it is worth establishing what has already been publicly confirmed.
Expedia has a live app inside ChatGPT. The Expedia page describing the integration states it directly: start planning in ChatGPT and keep the conversation going with the Expedia app. Dynamic flight and lodging results powered by Expedia, including pricing, availability, and rich visuals, surface directly within ChatGPT for a more complete and seamless way to plan.
Booking.com is also a named OpenAI travel partner. OpenAI’s own site carries a Booking.com customer story describing how Booking.com uses OpenAI models for smarter search, support, and discovery-phase travel experiences. The mechanics differ from Expedia’s live inventory integration. The direction is the same: major travel intermediaries are building directly into the AI-assisted planning layer.
Expedia is not merely another website outside the AI experience. It has a live path into the ChatGPT travel-planning interface. When a traveler’s personalized query surfaces hotel options, Expedia can supply live inventory, pricing, and a direct path to booking, without the traveler ever leaving the conversation.
Independent luxury hotels are not in that interface. They are not an option on the supply side of that integration. They are not competing at the level where Expedia is playing.
Two Layers. Both Point the Same Direction.
The commercial integration layer is the most visible problem. But there is a second layer underneath it, and independent hotels face it regardless of whether any commercial integration exists.
The commercial layer: Expedia and Booking.com are both publicly connected to OpenAI’s travel-planning ecosystem, with Expedia already offering dynamic travel results directly inside ChatGPT. When a traveler’s personalized query triggers hotel options, those options can come from platforms already integrated into the ChatGPT experience. Independent hotels with no direct presence in that interface have no seat at that table.
One clarification worth making: even if Expedia handles the booking layer, the informational selection layer still matters. When ChatGPT queries available inventory for a specific traveler profile, the properties that get surfaced and recommended from within that pool are still shaped by what the AI understands about those properties. Owning the informational layer does not bypass the Expedia interface. But it can influence whether a property is legible, distinctive, and matchable when AI systems evaluate available hotel options.
The informational layer: Even absent the commercial integration, independent hotels face a structural disadvantage. AI systems can draw from training data, retrieval sources, connected apps, and public-web material. Across that environment, intermediary descriptions often have more structure, repetition, and machine accessibility than the hotel’s own positioning. For many independent properties, the most repeated, structured, and easily retrievable public records of the property are not owned pages: OTA listings, aggregator summaries, Google hotel panels, review platform profiles, and metasearch descriptions.
When an AI system reconstructs what a hotel is, to match against a traveler’s specific profile, it works with what is available and accessible. The hotel that has published a canonical, structured, owned description of who it is and who its guest is has more control over that reconstruction than the hotel whose identity exists primarily as an aggregation of intermediary copy.
Both layers create the same strategic exposure. The commercial integration layer is documented. The informational layer follows from how AI systems encounter and process publicly available content.
The Personalization Update Changes the Stakes
The May 2026 ChatGPT update matters because it changes what the recommendation is based on.
Previously, a traveler asking for hotel suggestions would receive a general response based on their stated preferences in the moment. Now, ChatGPT can pull from the traveler’s Gmail: flight confirmations, hotel stays, loyalty affiliations, past trips. It can reference prior conversations. It can build a detailed picture of who this traveler is and what kind of experience they are actually looking for.
That makes the recommendation more specific, not more general. And more specific recommendations require more specific hotel identities to match against.
A traveler whose Gmail contains three years of boutique property stays, vineyard visits, and farm-to-table restaurant reservations is going to generate a different personalized profile than a traveler whose Gmail shows chain hotel loyalty points and conference bookings. The AI has enough personal context to look for something precise.
If the hotel’s available description, across every platform and source the AI can draw from, is generic luxury category language, it will not match a specific traveler profile with any precision. It will blend into the category. It will not be the answer.
If the hotel has built a distinct, owned description of who it is and who its guest is, in language that accurately reflects what makes the property specifically right for a specific kind of traveler, it has a better chance of being in the answer. Not a guaranteed chance. But a meaningfully better one than a property whose public identity is a compressed OTA listing.
This Is Not Presented as a Search Results Problem
The instinct in the hotel industry is to treat every AI development as a search problem. Where does the property rank? What keywords does it need? How does it optimize for the new algorithm?
ChatGPT’s personalized recommendation layer does not present ranking in the familiar search-results format. It generates a response. The traveler asking “where should I travel next based on what you know about me” is not going to see a list of ten hotels with star ratings and review counts sorted by score. They are going to get an answer, constructed from personal context, trained knowledge, and integrated supply.
There is no position one to bid for in the organic sense. There is no listing to update in the conventional sense.
What exists is a channel that Expedia entered with direct integration, and a content environment where intermediary descriptions have significant structural presence that most independent hotels have not matched with owned content. Neither of those problems is solved by keyword strategy or listing optimization.
The Pre-OTA Demand Problem Has a New Layer
The demand origin argument has always focused on the booking layer: the moment when a traveler is comparing properties and the OTA captures the commission. Most hotel marketing strategy is designed to intervene there: direct booking incentives, loyalty programs, rate parity negotiations.
ChatGPT’s personalized recommendation layer sits above all of that. It is the moment before the traveler has opened any booking platform. The moment when a traveler, based on everything an AI system knows about them, is told what to do next.
If Expedia is already integrated into that conversation with live inventory, the booking surface may exist inside the AI conversation itself. The comparison may have already happened before the traveler opened a browser tab.
A hotel that is accurately and specifically represented in that layer, whether commercially through direct AI integration or informationally through owned canonical content, has influenced the moment before the search began. A hotel that is absent from both layers has lost the demand origin moment before the traveler ever left their couch.
The Only Controllable Pre-Recommendation Lever
Independent luxury hotels cannot negotiate their way into Expedia’s ChatGPT integration overnight. They cannot rewrite the decade of aggregator copy that defines their public identity.
What they can control is what they publish from this point forward, and whether that content is structured for machine accessibility or written only for human readers.
Content built for AI accessibility, the discipline Knowledge Formation Optimization governs, means published, owned descriptions of who the property is, who the guest is, what makes the stay distinctive, and what kind of traveler the property is right for. It means ensuring that AI systems drawing from public-web material encounter owned, structured, canonical descriptions rather than leaving intermediary profiles as the clearest available record of the property. It means building the informational layer that the commercial integration layer cannot fully replace, because not every AI recommendation will route through an Expedia integration, and not every traveler will be in a context where the commercial layer is even active.
Hotels that built this before ChatGPT’s personalized recommendation engine became mainstream are better positioned. Hotels that have not have a narrowing window to change the informational layer, even if the commercial integration layer is already captured by intermediaries.
The Question Worth Asking
Every independent luxury hotel should ask one question: when a qualified traveler asks ChatGPT for a personalized recommendation, and the system draws from its trained knowledge, the user’s personal context, and available integrated supply: what does it find when it looks for your property?
Does it find Expedia’s version of your property? Or yours?
Does it find generic luxury category language that could describe any comparable resort in your competitive set? Or the specific, owned, distinctive language that makes your property the right answer for a specific kind of traveler?
Expedia already answered that question for its inventory. It is in the interface. It has live pricing and live booking. It was in the room before most hotels knew the room existed.
The informational layer is still open. But not indefinitely.
