Expert Views (15)

A very pertinent question, and your findings match ours almost exactly.

Going into the Boutique Hotel AI Index, we were conditioned to believe the hotels AI recommends most often would be the technically better-built ones, or at least weighted towards them. They were not. The most visible hotels looked much like the least visible: similar speed, similar schema (or lack thereof!), similar build.

Ask for hotel recommendations twice and you really do get two different answers. The best-known hotels appear more often, but the list changes even when you ask an hour apart. There is no ranking to win, only a likelihood of being mentioned to raise.

What raises it? Mostly what others write about a hotel: press, guides and reviews. On its own website, there is one main lever: consistently publishing useful content that reflects its ethos and its understanding of its location and its guests.

Optimise for the engines a little, but as with good SEO, write for the guest first and the machines will tend to follow.

So yes, I believe our industry is overestimating technical GEO. The one exception is agentic bookings. When AI agents book on a guest's behalf, technical readiness will matter, but that is readiness, not visibility.

Partly yes, and the bigger blind spot is measurement. A model quotes pages that answer a traveller's question, so useful content does more than schema alone.

Judging either kind of work is harder. SparkToro found that ChatGPT and Google's AI rarely give the same list of brands twice (less than 1 in 100), so a single screenshot proves little. The useful measure is how often a hotel is named across many repeated prompts.

Clicks are the easier part. Apart from a week in October when the tag went missing, ChatGPT marks its links with utm_source=chatgpt.com, so those visits show in GA4. An answer that names the hotel without a click leaves no trace there. On its Q2 2026 call, Booking Holdings' CFO put traffic from large language models at well under 1% of the group's room nights, so hotels still have time to prepare.

The first thing a hotel should do now is check that robots rules and JavaScript-only pages do not keep AI crawlers out. The second is to pick a fixed set of traveller prompts and track how often the hotel is named, with Semrush, Ahrefs, Peec AI or by hand, before and after publishing new content.

Technical optimization is more important than ever. We are seeing this everywhere, reflected in data in various studies. Editorial articles are actually structured very well. They are used as third-party trust factors by AI algorithms to verify, find, and recommend hotels. That's why you are seeing them pop up in search results.

When it comes to online content enabling your brand or experiences to be found, the key word here is: useful. Are we underestimating the value of consistenly publishing useful content? Perhaps.

But more importantly, I think a vast majority of travel brands just focus on consistenly publishing content. Not necessarily useful, and to be honest, more often than not, purely promotional.

So are we overestimating technical GEO? Just like we were overestimating technical SEO, I'd say. Don't get me wrong: I agree with Simone. One should look into schema markup, structured data, FAQ and website optimisation.

Alas, when it comes to GEO, a lot of the influence (and citations) stems from digital outposts outside your control, on third party sites, OTA and influencer blogs and social platforms.

So technical GEO is important, but really needs to be looked at in conjunction with technical SEO, including Google business profiles, social media contents and third-party platforms for a more holistic approach.

Interesting that this should come up as a topic for discussion.

Let's face it. Noone really can clearly explain how AI works or why it makes the choices it makes. It starts there.

My experiences are similar. Ask the same question - even of the same engine - get a variation of answers on that same topic. Formal tests support the same outcome.

Although, when we consider that if we truly claim 'artificial intelligence', then like a discussion with a person, given the moment, time, background and incremental input impacting their view on a topic, would they also have a slightly different answer?

Perhaps we are not considering the sheer scale of information that the source has available and the speed at which even small changes might impact the reasoning from one hour to the next. Can we truly fathom the moving parts that could nuance the response?

It may be the case that we seek a consistency or certainty that is no more realistic from AI, than from having a discussion with a human being.

Maybe that is a good thing. Maybe that is how things should be. Maybe that is what makes a slightly more level playing field.......

Should there be only one standard outcome?

Are we overestimating technical GEO? We are, if we expect schema and structure to get a hotel mentioned.

AI search has splintered SEO into new acronyms, and GEO (generative) is where hotels have the biggest opportunity. It reads what it finds online and then writes the answer itself. Ask twice and the answer changes, because there is no ranking to achieve. It writes a fresh one each time from the material it trusts most.

What I like best is that it's forcing marketing and content to be more useful. The content that gets surfaced is the one that helps a guest decide, which puts the guest back at the center, where they've always belonged.

These engines are "travel agents" after all, and a guest asking an open question wants a conversational answer with real context behind it. It makes sense why content marketing, blog articles, and storytelling keep surfacing.

These AI systems look for straight facts about a property, paired with helpful, trustworthy content on your website and across the web that it can validate with reviews. The engine's job is to hand its user a confident answer based on its findings.

Site speed and markup still matter, but usefulness is the optimization.

Referrals from the AI platforms are accelerating and hoteliers should not ignore them. In the past, travel discovery was dominated by Google. Today, increasingly so, by the AI platforms and Google AI Mode. 

So, what should hoteliers do to compete for mentions, referrals and ultimately - direct bookings from the AI platforms?

a) invest in GEO (Generative Engine Optimization) to have your property featured in the AI trip planning recommendations on ChatGPT, Perplexity, Microsoft Copilot, Claude, and Google AI Overviews. The goal here is to create content that is clear, credible and authoritative to be included in citations and recommendations

b) invest in AEO (Answer Engine Optimization) to get your property’s unique value proposition featured in the answer boxes (ex. Google’s Gemini). AEO requires structured content that the search engines can extract and display as direct answers. 

c) invest in SEO - Even today, 70–80% of SEO best practices that worked for decades also work for GEO and AEO. 

d) invest in content marketing and social media to “earn recognition” on the AI platforms in the form of mentions, citations and recommendations. 

e) invest in Online Reputation Management (ORM) - customer reviews are an important source for brand authority in the eyes of the AI platforms

Technical factors are a baseline, not a differentiator for a competitive advantage.

Of course you need schema, structured data, fast sites, indexable content and ease of navigation, etc. But, these have been around for YEARS for SEO purposes.

If a hotel is lacking these basics, then they will just be rarely found, depending on the market competitors.

Research supports universal AI rankings to be almost impossible due to variations in how LLMs (continue to) learn, variations in prompts, personalization, variations in responses. (See SparkToro research Jan 2026)

Yes, the industry is overestimating the importance as a part of a value proposition over useful content, written for HUMANS and AI.

We may be overestimating how much technical GEO can accomplish without giving AI something useful to find.

Schema and structured data can help make information accessible. But an FAQ answering a real traveler’s question gives that accessibility a purpose. Hotels should spend at least as much time deciding what questions to answer as worrying about how to label the answers.

I also suspect changing recommendations are here to stay. Variation alone doesn’t prove an AI is learning, but as these systems get better at retrieving, combining and presenting information, then incorporating signals about whether their answers satisfy users, we should expect their recommendations to evolve. A permanently fixed answer would hardly be the goal.

Meanwhile, we’re all participating in a digital marketing arms race, continually increasing, and hopefully improving, the information these systems must sift through. Today’s visibility tactic can quickly become tomorrow’s background noise.

Technical foundations matter. But the more durable investment is consistently publishing clear, accurate, useful information that helps someone make a decision.

We should be asking less, “How do we get AI to mention our hotel?” and more, “What have we published that makes our hotel worth mentioning?”

I’d push back gently on the framing. The real risk isn’t overestimating technical GEO. It’s assuming either camp has found a lever that works reliably.

If the same model answers the same question differently twice, then no one, whether technician or content publisher, is optimising for a stable ranking. We’re all optimising for probability. That changes the job from “rank first” to “be present, repeatedly, across the questions that matter.”

And presence is earned by being the most useful source, not the most structured one. Schema helps a machine parse you. It doesn’t give it a reason to cite you. Your finding that FAQs and editorial content dominate retrieval fits what I see commercially: AI rewards pages that sound like the buyer’s question, not the brand’s pitch.

Here is the angle I rarely see discussed. The best source material in a hotel isn’t in the marketing calendar. It’s in the sales office: the RFP clarifications, the corporate travel manager’s objections, the MICE planner’s recurring questions. That is first-hand, specific, and impossible for a competitor to replicate.

Yes, the hotel industry is focusing too much on technical web fixes and not enough on helpful content.

AI search tools don't recommend a hotel just because its website code is tidy. They look for direct, detailed answers to real traveller questions—the kind found in blog posts, travel guides, and FAQs. Because AI doesn't give the exact same answer every time, success isn't about ranking #1 on a fixed list; it's about consistently being featured across the web as a trusted answer to what travellers are actually asking.

As a strategy, yes. As hygiene, no.

SEO was a fight for a spot on a list. AI doesn’t do lists. It gives you one answer, maybe two. Nobody scrolls to page two of a recommendation.

As for GEO, we’ve seen this film before. First a hack, then a craft, then an ad budget. Then the answer box is for sale again, just with a new landlord.

And here’s the catch: AI listens to consensus, not to your manifesto. It hears what guests, critics, guides and forums say about you. Your own channels are one voice in that choir. Tune them, by all means. They won’t carry the song.

“This changes everything. Again.” That was Apple’s line for the iPhone 4. Our species has said the same about every moment since the printing press. We’re usually right that it changes. We’re almost always wrong about how.

So clean up your act: your data, your website, your booking path. Then ask the only question that matters: why would anyone, human or machine, recommend you with conviction? No prompt or plugin can answer that. A product worth talking about can, and so can the people who talk about it.

Technical optimization is critical in GEO, but it should support content quality rather than become the primary objective because content is still king.

As travelers increasingly rely on AI platforms for travel planning, hotels must understand what questions consumers are asking and provide relevant answers. For example, instead of simply listing hotel amenities, a property could publish rich textual and visual content explaining what makes its location unique, which nearby attractions are worth visiting, or how its services accommodate different travelers' needs. Such content can help AI systems understand a property's distinctive attributes and potentially recommend it in relevant conversations.

Meanwhile, these rich narratives should be supported by structured data, accessible content, and consistent property information to help AI systems discover and interpret hotel information accurately. Technical sophistication alone cannot compensate for content that offers little value to travelers.

Ultimately, GEO is not a competition between technology and content. Hotels should focus on creating authentic, informative, and machine-readable content that serves both human travelers and AI agents. The goal is not simply visibility, but credibility and recommendability.

Very possibly--I think we’re giving GEO the starring role when it’s a supporting character. The star that reliably gets left out is the consumer’s intent.

The consumer’s agent executes the search with some understanding of the traveler’s preferences and requirements for this particular trip—information that no hotel, OTA, or other participant in the booking chain necessarily has. Its job is to find the properties best suited to those needs and gather credible evidence of that fit--partially through GEO.

If I want an intimate, romantic, independent hotel, your branded mega-property is unlikely to make the shortlist, however immaculate its schema markup. There are limits to what good punctuation (and copy) can accomplish.

Obviously, hotels should make their information readily accessible and understandable to agents, and preferably through MCP. Solid property content, guest reviews, editorial coverage, and social discussions can all help establish suitability. But their value on this trip depends on what the traveler is looking for.

The question I’d ask is: “Can the agent establish that we’re a strong fit for this customer’s trip?” GEO can help it find and interpret the evidence. The hotel still has to suit the customer. Personalization comes next, but that's another discussion.

I’ve been thinking that, with all the attention around GEO, we may be at risk of repeating the same approach we’ve taken with SEO for years. The tools are changing, the intermediaries are changing, but we’re still largely focused on how to get found.

Structured data, Schema.org, machine-readable content, interoperability protocols. All important, of course. And many destinations still have a lot of work to do on these fundamentals.

But are we sure the main challenge is technical?

Making a destination understandable to an AI system is one thing. Making that destination emerge as a relevant recommendation for someone, based on their motivations, is quite another.

And this brings us back, I think, to questions that are far from new, but much more strategic.

Why should someone choose that particular place? Who does it matter to? What desires, expectations or sense of belonging does it connect with? And what makes it stand out from dozens of alternatives?

In tourism, we’ve invested heavily in visibility, often much less in building a distinctive identity and a consistent reputation over time. And now we risk simply shifting that obsession from Google’s first page to AI-generated answers.