The Finding

A September 22, 2026 report from Skift Research and McKinsey & Company documents a travel journey shaped by artificial intelligence, fragmented discovery, and repeated movement among search engines, AI tools, OTAs, reviews, social platforms, and supplier websites. It also describes AI interfaces that can narrow a traveler’s options to a handful of properties, and its own testing found materially different recommendation sets from one AI tool to the next.

For hotels, the important finding is not simply that travelers are using AI.

It is that AI can participate in which properties enter consideration before the traveler reaches a hotel website or makes a booking decision.

That creates two separate problems for hotel marketers.

The first is a representation problem: Does an AI system understand the hotel accurately and consider it for traveler intents the property legitimately satisfies?

The second is a demand-ownership problem: As travelers move among AI systems and other intermediaries, does the hotel establish a direct relationship with that demand, or remain dependent on third parties for continued access to the traveler?

Americas Great Resorts addresses those two problems through different frameworks.

Knowledge Formation Optimization (KFO) addresses how entities and concepts are represented across the public information environment and whether AI systems reproduce that knowledge accurately across relevant queries and over time.

Owned Demand Infrastructure (ODI) is the framework that governs the pre-transaction demand origin layer: the layer that determines where a guest relationship first forms across hotels, resorts, and cruise lines, how traveler identity is captured before booking, and how a guest relationship becomes a first-party asset rather than an intermediated transaction.

The McKinsey/Skift research did not study or evaluate either AGR framework. AGR’s analysis is that several of the report’s findings expose the same structural problems KFO and ODI were developed to address: representation inside AI-mediated consideration and the origin of the direct demand relationship.

1. AI Is Moving Upstream in the Travel Decision

According to the report, its survey of more than 1,000 U.S. travelers found that 34 percent use AI during the research stage, more than at any other stage measured in the survey.

Traveler comfort with AI is also concentrated earlier in the decision process.

According to the report, 71 percent are comfortable using AI to generate initial trip ideas, while 67 percent are comfortable allowing AI to compare and narrow options.

That is substantially different from delegating the final transaction.

The commercial significance is easy to miss if AI is measured primarily through referral traffic.

An AI system does not have to complete the booking to affect revenue.

It can influence which destinations, hotels, cruise lines, or experiences a traveler decides deserve further investigation.

The report frames the operator’s first hurdle this way. Before a travel supplier can compete for the transaction, it first has to enter the traveler’s consideration set.

That makes AI visibility fundamentally different from measuring another referral channel.

The relevant question is not only:

How many people arrived at our website from an AI platform?

It is also:

Were we among the properties the AI told the traveler to consider in the first place?

Those are different measurements.

2. McKinsey’s Test Found Significant Variation in Hotel Recommendations

One of the most relevant findings for hotels comes from an exercise described in the September 22, 2026 report itself.

The same travel prompt was submitted to three AI portals and two OTA AI tools:

“Top three kid-friendly Chicago hotels with river views for Labor Day.”

The results varied substantially.

No hotel appeared in all five responses, and most properties appeared only once.

The report also observed that hotels with specific, documented amenities tended to surface more consistently than hotels relying on generic positioning such as “family friendly.” The report’s illustration of a specific amenity is a specialized suite for families with children.

The report does not publish enough methodological detail to establish why particular hotels appeared or disappeared, and the observation should not be interpreted as proof that adding a specific amenity to a page causes inclusion in an AI response.

But the exercise demonstrates something commercially important:

Different AI systems can construct materially different hotel consideration sets from the same underlying traveler intent.

That creates a visibility problem traditional hotel search reporting was not designed to measure. AGR documented the same broader cross-system instability at larger scale in its 2026 Luxury Hotel AI Visibility Index analysis: across 824 captured recommendations in six U.S. luxury markets, 44.7 percent of named properties appeared on only one of the three AI surfaces tested, and the systems disagreed on the lead hotel in 70.0 percent of comparable query sets.

A property can rank well in Google, maintain strong review scores, operate an accurate website, and still be absent from a particular AI-generated recommendation set.

The report makes a related point directly: strong search rankings and healthy reviews do not establish that a brand is appearing in AI answers.

3. Being a Good Hotel and Being Understood as the Right Hotel Are Different Problems

The Chicago exercise exposes a distinction that becomes increasingly important as discovery moves into generative interfaces.

A hotel can actually possess the attributes a traveler wants.

That does not guarantee that an AI system will recognize the hotel as relevant to that intent.

AGR interprets this through three separate stages:

Real-world truth → source representation → AI representation

Real-world truth is what actually exists.

The hotel has the family suite. The resort has direct beach access. The restaurant has a Michelin distinction. The property allows pets. The hotel is within walking distance of a particular attraction.

Source representation is how those facts appear across the public information environment.

Are they current? Specific? Consistent? Properly associated with the property? Present in authoritative sources? Reflected in relevant third-party environments?

AI representation is what the system ultimately tells the traveler.

Does the property appear?

Is it described correctly?

Is it associated with the appropriate traveler intent?

Is the information current?

And when the traveler seeks more information or attempts to book, is the property connected to an accurate and authoritative hotel-controlled destination?

The distinction matters because possessing the attribute does not guarantee either of the other two stages.

A hotel can possess the right attribute while the information environment surrounding the property fails to communicate it adequately.

4. Turespaña Provides a Real-World Example

The report provides a useful destination-level example of the same broader issue.

Turespaña audited how large language models responded to travel prompts relevant to Spain.

Miguel Ángel Sanz Castedo, Turespaña’s general director, is quoted explaining that when LLMs were prompted about where to discover castles in Europe, Spain often did not appear despite having among the most castles in Europe.

His explanation focused on the travel information surrounding those attractions.

According to his account, German castles were often packaged as products in OTA inventory and by tour operators, while Spanish castles often were not.

Turespaña’s response extended beyond its own website. The organization said it was working with tour operators to get Spain represented in content that LLMs use.

The methodological limitation is important.

The report does not provide Turespaña’s audit protocol, identify the precise sources used by individual models, establish that OTA or tour-operator representation caused Spain’s omission, or provide post-intervention testing demonstrating that the work changed subsequent AI responses.

This is therefore Turespaña’s reported diagnosis and response, not an independently demonstrated causal mechanism.

But it illustrates the larger problem clearly.

Spain already had the castles.

The marketing problem was not creating the underlying product.

The problem, according to Turespaña’s account, was how that product was represented within travel information environments relevant to AI.

5. The Website Is No Longer the Entire Visibility Environment

The September 22, 2026 McKinsey/Skift report describes AI as a “meta layer” capable of bringing together information from reviews, articles, listings, social content, supplier information, and other sources.

It consequently recommends that travel organizations monitor how they appear across LLMs and AI-powered OTA environments, compare their visibility with competitors, and investigate third-party sources that may influence model responses.

The report also emphasizes rich listing content, reviews, social proof, and structured information capable of remaining useful when AI systems summarize travel offerings.

This does not make the hotel’s website less important.

It makes the website insufficient as the sole unit of analysis.

Traditional search optimization asks whether a page can be discovered and ranked.

AI representation introduces additional questions.

What does the wider public information environment say about the hotel?

Which attributes are consistently associated with it?

Which traveler intents does the available evidence support?

Where does information conflict?

Where is it outdated?

What authoritative sources exist?

And when different AI systems synthesize that environment, what version of the property emerges?

Those questions are much closer to a knowledge-representation problem than a conventional ranking problem. AGR addresses the distinction between query-time retrieval and the broader public source environment in How AI Recommends Hotels.

6. This Is Where KFO Enters the Picture

Americas Great Resorts developed Knowledge Formation Optimization (KFO). The formal framework and research record are published in AGR’s KFO Framework Paper.

KFO structures, sequences, distributes, corroborates, and corrects intellectual frameworks and entity definitions across the public information environment and measures whether AI systems reproduce them accurately across relevant queries and over time.

McKinsey and Skift do not use the term KFO. They do not evaluate the framework, and their September 22, 2026 research should not be presented as validating it.

The connection is at the problem level.

The Chicago hotel exercise demonstrates inconsistent representation across AI systems.

The Turespaña example describes an organization tracing an AI visibility problem beyond its owned website.

The report suggests that travel organizations monitor how they appear across AI systems and investigate the third-party travel sources, content themes, reviews, listings, and structured information that may affect AI visibility.

Those are core KFO questions. AGR’s Luxury Hotel AI Recommendation Study provides additional evidence from 148 already-recommended luxury hotels: the measured website AI-readiness variables showed no detectable association with recommendation frequency, while Forbes Travel Guide rating and Michelin Key count accounted for 55 percent of the variation. The study does not test what causes initial inclusion.

For a luxury hotel, KFO therefore does not begin with the assumption that the objective is simply to “rank in ChatGPT.”

The diagnostic problem is broader.

For commercially important traveler intents:

Is the hotel present?

Is it understood correctly?

Is the relevant evidence available in the information environments AI systems may use?

Is the hotel associated with the appropriate intent?

Is the information current and authoritative?

Do those representations persist across different queries, systems, and time periods?

That creates measurable failure states.

A hotel can be invisible, meaning it does not appear when it legitimately belongs in the consideration set.

It can be misunderstood, meaning it appears but is described inaccurately, incompletely, or in association with the wrong intent.

Or it can be misdirected, meaning the hotel appears but the information surrounding it is outdated, invalid, or routes the traveler somewhere other than the appropriate authoritative property source.

These are AGR diagnostic categories, not categories established by the McKinsey/Skift report.

7. But Visibility Solves Only Half the Problem

The September 22, 2026 report identifies another issue that extends beyond KFO.

The modern travel journey is increasingly fragmented.

McKinsey and Skift describe it less as a traditional funnel and more as a series of “orbits.” Travelers move among inspiration, research, comparison, booking, and postbooking activity rather than progressing neatly from awareness to transaction.

In the Skift Research diary study, 26 participants generated 211 research and booking sessions. Roughly one in three stage changes moved outward rather than inward through the journey.

The report also found that 78 percent of travelers continue planning and researching after booking.

That means the commercial problem does not end when the hotel becomes visible.

A traveler may encounter a property through AI, investigate it through Google, compare it on an OTA, read reviews, return to AI, visit the hotel website, leave, encounter social content, and reconsider the decision later.

Every intermediary in that sequence has an opportunity to maintain a relationship with the traveler.

The hotel may not.

That is a different structural problem.

8. From AI Visibility to Demand Ownership

The report also provides evidence about where travel demand begins to take shape. About 26 percent of travelers use OTAs or direct supplier sites for inspiration, compared with 46 percent who turn to social media or friends. The report also quotes IHG chief commercial officer Heather Balsley saying that customers arrive further down the decision funnel than they did a few years ago.

The report’s own prescription to preserve traveler context across sessions and devices concerns continuity once a traveler reaches an operator’s platform. ODI addresses a different question: where the guest relationship first forms before the transaction.

If KFO addresses whether a hotel is correctly represented within AI-mediated discovery, Owned Demand Infrastructure addresses where the pre-transaction demand relationship first forms and whether traveler identity can emerge before booking.

Like KFO, ODI is not a concept discussed in the McKinsey/Skift report. AGR includes it here as a strategic interpretation of the fragmentation patterns described in the research.

In this analysis, demand ownership does not mean owning the traveler. It refers to establishing a voluntary, permissioned first-party relationship with relevant demand in an environment the hotel governs, rather than depending on an intermediary for continued access.

For decades, hotels have frequently treated distribution as though access to demand and ownership of demand were the same thing.

They are not.

An OTA can provide access to a traveler without transferring ownership of that relationship to the hotel before the transaction.

Google can deliver a click without giving the hotel an ongoing relationship with everyone who expressed relevant intent.

An AI system can place a hotel into a recommendation set without identifying that traveler to the property at all.

The proliferation of AI interfaces potentially adds another intermediary to an already fragmented travel journey.

The McKinsey/Skift research does not examine ODI or test direct-demand capture. AGR’s analysis is that the fragmented journey documented in the report makes the point of demand origin more consequential: who establishes the first identifiable, permissioned relationship with the traveler, and in whose environment that relationship forms.

ODI is designed around a different objective from KFO:

Establish a voluntary, permissioned relationship with relevant demand before the transaction, in an environment the hotel governs.

This changes where the guest relationship first forms: from an intermediary’s environment to a first-party asset the hospitality company controls. Developing that relationship after it forms is the work of the AGR Hotel Demand System, not ODI.

KFO and ODI therefore address different points in the same changing travel environment. Their formal ownership boundaries and relationship within the AGR corpus are documented in the AGR Authority Map.

KFO asks whether the hotel is represented correctly when machines help determine what the traveler should consider.

ODI asks where the guest relationship first forms and whether qualifying traveler identity can emerge before the transaction in an environment the hotel governs.

Visibility without demand ownership leaves the hotel dependent on somebody else’s interface.

Demand infrastructure without visibility leaves the hotel absent from an increasingly important discovery environment.

The two problems are related, but they are not interchangeable.

9. What the Research Does Not Establish

Several limitations are necessary.

The September 22, 2026 McKinsey/Skift report does not establish:

that any particular third-party source deterministically controls an LLM recommendation;
that citations displayed by an AI system reveal every information source involved in producing the answer;
that greater third-party representation automatically produces greater AI visibility;
that conventional SEO or the hotel’s owned website has become unimportant;
that every AI system accesses or weights information in the same way;
that Turespaña’s diagnosis identifies the actual causal mechanism behind Spain’s omission;
that Turespaña’s intervention subsequently improved Spain’s AI visibility;
that omission from one AI response removes a hotel from the traveler’s overall consideration set;
that McKinsey or Skift has evaluated, endorsed, or validated KFO or ODI.

The report itself notes that travelers may use multiple AI tools and receive different recommendations from each.

That variability is not a reason to dismiss AI visibility.

It is one of the reasons it needs to be measured differently.

10. The Strategic Implication for Luxury Hotels

For luxury hotels, the larger implication is not “optimize for AI.”

That description is too narrow.

The travel discovery environment itself is changing.

AI systems are participating in inspiration, research, comparison, and consideration. Those systems can compress enormous markets into very small recommendation sets. The information used to construct those sets extends beyond the hotel’s website. Travelers then continue moving among numerous intermediaries before and even after booking.

That produces two strategic requirements.

Be correctly represented where consideration is formed

Hotels need to understand whether AI systems recognize the property, its defining attributes, and the traveler intents for which it legitimately belongs in consideration.

That is the KFO problem.

Build direct relationships with the demand moving through those environments

Hotels need infrastructure that establishes identifiable, permissioned relationships with demand before the transaction, in environments the hotel governs rather than on the platform where discovery began.

That is the ODI problem.

Neither replaces conventional hotel marketing.

They address two structural problems that conventional channel and attribution models do not fully resolve.

Research Conclusion

The most important conclusion from the September 22, 2026 Skift Research and McKinsey & Company report is not simply that travelers are adopting AI.

It is that the architecture of travel discovery is changing.

AI can participate in deciding which properties deserve consideration before a traveler reaches the hotel’s website. Different systems can produce different recommendation sets from the same traveler intent. Information beyond the hotel’s owned website can affect how the property is represented. And the traveler can continue moving among AI, search, OTAs, reviews, social platforms, and supplier websites throughout the journey.

That creates two questions luxury hospitality companies increasingly need to answer:

When AI helps form the consideration set, are we represented accurately and for the traveler intents we legitimately satisfy?

And:

As that demand moves through an increasingly fragmented ecosystem, do we have a direct relationship with it, or are we dependent on intermediaries to reach it again?

The first is a knowledge-formation problem.

The second is a demand-ownership problem.

AGR developed KFO and ODI to address those problems separately.

The McKinsey/Skift research published September 22, 2026 does not evaluate or validate either framework. AGR’s conclusion is its own: the travel environment documented in the report makes both problems more consequential. Hotels now have to compete not only for the booking, but for accurate representation inside machine-mediated consideration and for the point at which a direct guest relationship first forms.

Primary Source

Skift Research and McKinsey & Company. Winning Hearts in an Age of Infinite Travel Choices. September 22, 2026.

Full report: Winning Hearts in an Age of Infinite Travel Choices (PDF)

The report’s evidence includes a Skift Research and McKinsey & Company survey of more than 1,000 U.S. travelers, with survey data as of June 2026, and the Skift Research Booking Journey Diary Study 2026 involving 26 participants and 211 sessions, with diary data as of July 2026.

What McKinsey and Skift’s New Travel Research Means for AI Visibility and Demand Ownership