The Citation You Can't Cash: 14 Studies on How Visitors Decide Now, and How DMOs Prove Its Part.

After a whirlwind of travel from ESTO in Philadelphia and Destinations Texas, I’m finally back at it, squirreling away the latest data in our world of travel.

And it’s safe to say it’s technically fall/pumpkin spice season in my neck of the woods at a 93 degree temp in Texas, which means it’s the season for those end-of-summer agency reports with reach, CPMs, impressions, and engagement rates.

Let’s be real: I know what you do with those reports.

You skim them, you forward them to your marketing team, and then you open your STR dashboard to see the numbers you are judged on at the end of the day. Occupancy, ADR, RevPAR.

You pull up KeyData to check your short-term rental performance. You check whether your hotel partners are happy, whether your city council is asking hard questions, and whether the dollars are flowing into your local economy.

That is the real job. And as a marketer, I respect that. It’s the end game.

But I’m writing to you today because there is a structural shift happening underneath all of those metrics, and the dashboards you rely on are not catching it yet. The ground is moving more than it ever has, and the data from the last few months makes it impossible to ignore.

I had the opportunity to sit in on Robert Patterson’s session at ESTO, where he walked through the new AI Readiness Report from MMGY Travel Intelligence, Mindtrip, Sabre and Sojern. It surveyed 1,006 U.S. leisure travelers and 103 destination marketing leaders.

The headline number?

The travel industry is operating about two full years behind the visitors it exists to serve.

Yet, AI was on nearly every agenda in every session. And likely most conference sessions this year. Even for an AI dork like me, it was a lot.

And one thing I keep realizing is that a LOT of the AI talk on the conference circuit tends to be disconnected from the operational reality of running a destination organization with a small budget, a stretched team, and a board that wants heads in beds. The panelists are right about the direction of travel. The gap between the keynote stage and your Monday morning staff meeting is, well, a lot. Let’s talk about it.

What I want to do here is give you the unvarnished data, pulled from as many independent sources as I could get my hands on. I made myself one rule for this issue: no single study gets to carry the whole argument. You’ll see Longwoods, Similarweb, Sojern, Amadeus, Expedia, Google, Cloudflare, CoStar, KeyData, AirDNA, Zartico, Gallup, and a few independent researchers who are doing the needed work. Every number is linked to the people who measured it. The data is U.S. wherever a U.S. read exists, the handful of global or multi-country figures are labeled as such, and where a number is older or thinner than I’d like, I say so. Hope there’s some noise we can cut through on a Friday afternoon.

Let’s get into the goods.

The Visitor Moved, the Dashboard Didn’t.

The gap between what visitors expect and what destinations can deliver did not open because we stood still. Most DMOs I talk to are working harder than ever and are, frankly, spiraling. The problem is this rapid pace of change.

If we map this onto a classic technology adoption curve, visitors have blown past the early adopters and are barreling into the late majority. They are making AI a regular, habitual part of how they discover, plan, compare, and book trips. Meanwhile our industry as a whole is still in the early adopter phase, experimenting with chatbots, testing translation tools, generating content with the LLMs (please don’t replace storytelling this way), and still trying to figure out which capabilities deserve real investment.

The freshest independent read on visitor behavior comes from Longwoods International’s American Travel Sentiment Study. In Wave 109 (fielded July 7 to 9, n=1,000), 28% of American adults said they had used AI to plan a trip in the past six months, up from 19% in August 2024. Of the people who did, 54% used it to find attractions and activities, 44% for inspiration on places to visit, and 42% to find dining. Read that first one again. More than half of the people using AI for travel are using it to answer the exact question your destination website exists to answer.

Amadeus fielded 1,000 U.S. travelers in July and published the results August 3. 74% have used AI during the planning stage of a trip. 65% are interested in booking and paying directly inside an AI assistant. And only 20% would let an AI tool spend money on their behalf without additional approval. We’ll come back to this.

The Mindtrip, Sabre, Sojern and MMGY Travel Intelligence report lands in the same place from a different angle, and it deserves the credit for putting the two sides of the gap in one spot:

• 54% of leisure travelers have used AI to plan a trip, and 75% plan to on their next one.

• 63% would use an AI trip planner directly on a destination’s website if one existed.

• On the supply side, 22% of destination organizations offer one, and 26% have a documented AI strategy in their marketing plan.

Three-quarters of visitors intend to use AI for their next trip, and about a quarter of us have written down a strategy for it. The rest are running scattered pilots without a framework to turn them into lasting organizational capability. And kudos to those who have, because it’s a lot of trial and error with a moving target.

Sojern’s State of Destination Marketing report (350-plus DMOs worldwide, so this one is global, fielded with Dynata, and yes, Sojern sells advertising, so read it with that in mind) has a pair of numbers nobody has put side by side. 51% of DMOs say they are concerned about or actively preparing for AI-driven search disruption. In the same study, 93% say they are concerned about or preparing for AI in general. We are broadly anxious about AI and specifically under-anxious about the one part of it that changes how visitors find us.

It gets better. 64% of DMOs say they are now writing content specifically formatted for AI engines. 34% monitor whether their content shows up in AI results. And 14% measure AI search visibility as a KPI. We have moved to producing for the LLMs without instrumenting whether producing worked.

That is the whole story of this year in just three numbers.

And visitors are voting with their clicks. 25% say they are less likely to use a destination’s website at all if it does not offer AI features. You have spent years and real money building a beautiful, content-rich website. A quarter of your potential visitors are telling you they will not land on it.

And one more, Similarweb tracks visits to the AI assistants themselves, and ChatGPT’s share of generative AI website visits fell from about 76% to 53% between June 2025 and May 2026 (a worldwide figure), with Gemini climbing to roughly 27% and Claude to nearly 9%. If your team is checking how your destination shows up in ChatGPT alone, you were checking half the surface in May, and you’ll be checking less of it by December. Name the assistants. Check each one.

The existential awareness is there, even if the action is not. 72% of DMO leaders agree that organizations failing to adopt AI within two years risk becoming irrelevant in the destination discovery process.

You know the clock is ticking. The question is what to do about it.

The Citation You Can’t Cash

For twenty years the DMO’s job was to win the search engine results page. You optimized for keywords, you built backlinks, you published blog posts, and you drove traffic to a website where you controlled the narrative and the conversion path. That model is being hollowed out from the inside, and the hollowing has a specific shape that we haven’t yet figured out fully.

Similarweb published this on September 3. ChatGPT answers about travel included a web citation 22.6% of the time, against a 6.8% average across all conversations. AI referral traffic to travel sites grew 115.6% year over year, and ChatGPT alone sends more than 80% of AI referrals to the top 1,000 global domains. Adobe’s own analytics show AI traffic to U.S. travel sites up 2,215% between October 2024 and May 2026, off a base that was close to zero. (Adobe’s more detailed cut says those AI-referred visitors stay 70% longer and convert 28% less.)

Now the other half. Pew Research tracked the real browsing of 900 U.S. adults and found that when an AI summary appeared on Google, people clicked a traditional result in 8% of visits, against 15% when no summary appeared. They clicked a link inside the summary in only 1% of visits. One.

Then there’s who gets cited when a visitor asks an AI assistant about a destination. Search.Agency sent 100 unbranded travel prompts to Google AI Mode in July and classified every source it returned. Official tourism boards and government sites earned 1.2% of citations. Independent travel blogs and content sites earned 56.8%. That is one engine on one day with 100 prompts, so call it directional and nothing more. Directionally, this is an insane finding.

Okay now let’s put this together. Travel gets cited more than any other category. And almost nobody clicks the citation. And when the citation does exist, it is very rarely us. Super cool. A citation proves you exist in the answer. It does not prove anyone came, and it does not prove the answer sent them to your partners. Mindtrip’s own survey work in July found 70% of DMOs reporting organic search declines they attribute to AI-powered results, which is the same phenomenon seen from inside our own analytics.

This is a structural erosion of the primary discovery channel destination marketing has relied on for two decades, and the replacement channel does not send clicks. It sends decisions.

Stewardship, with the Credit Where it Where it Belongs

Dr. Jens Thraenhart, a Founding Partner and CEO of Chameleon Strategies and Second Vice Chair of the UN Tourism Affiliate Members Board wrote the essay in August called The DMO’s Job Is Not Promotion Anymore. It Is Data Stewardship. His argument is that an AI assistant does not discover your destination through a hero video. It discovers you through structured information it can read about what is open, what it costs, what it’s near, and what people who went there said about it. A destination that has not organized that information in a form the assistants can consume is simply absent from the dataset, and a third party fills the gap with whatever it can find. Agreed.

Miguel Sanz of Turespaña made the same case on stage with Rafat Ali and Fred Dixon in a Skift session published back in April that the new job is shaping what ChatGPT, Gemini and Claude recommend, on top of building a website or buying media. Jens goes one step further with an example I love for us. Hong Kong Polytechnic University now runs a tourist satisfaction index on more than 1.25 million TripAdvisor reviews across 13,694 providers, scored by an LLM and refreshed monthly. Think about your annual visitor satisfaction survey. Now imagine it running every month off what visitors already said, by district and by sector. That is a different category of instrument, and it exists today. It’s what I have been doing with my DMO clients through my founded company Zertura, because I truly believe its a monthly need, not a yearly.

Nicolas Sitter, an independent researcher, crawled 105,002 hotel websites this year across several countries, the U.S. included. 96.7% have zero AI-specific rules in their robots.txt file, which means almost no hotel has formed an opinion on which AI crawlers it lets in. In a companion study, 55.8% of hotels carry some JSON-LD structured data and 36.3% carry none at all. I looked for the equivalent study on destination websites and could not find one. Nobody has yet measured how many DMO sites carry the structured data that Jens is describing. (If you know of one or want to partner on making this happen, shoot me an email!)

And the window is moving without us. As of today, Cloudflare blocks mixed-use AI crawlers by default on pages that carry ads, meaning any crawler that will not tell a site owner whether it is there for search, for an agent, or for training. A lot of DMO sites sit behind Cloudflare. Your IT or web provider should know what your CDN and your robots.txt are telling ChatGPT, Claude, Perplexity and Google today, because an infrastructure vendor may have already decided for you.

Then look at what happened on August 27. Google switched on agentic hotel booking inside AI Mode with ten launch partners: five chains (Choice, Hilton, IHG, Marriott, Wyndham) and five OTAs (Booking.com, Expedia, Hotels.com, Priceline, Trip.com). No DMOs. No local boutique hotels. No attractions, no restaurants. When an AI assistant moves from answering to transacting, the entities wired into the transaction are the ones who own inventory. We own no inventory. That means the only lever a destination has in an agentic booking flow is upstream, at the moment the assistant decides which destination to name and which local partners to describe. Stewardship is what happens at that moment.

There is a silver lining here. Robert’s slides carried a figure that 34% of travelers trust AI recommendations more when they appear on an official destination website, rising to 55% among travelers who already use AI and confirmed by Expedia’s independent where 68% prefer to book with a trusted travel brand over an AI chatbot or agent. The DMO brand still carries weight. The authority is yours to claim, but only if you build the data infrastructure to support it. Schema helps AI understand a destination, and third-party validation helps AI trust it.

Inspiration VS. Utility

Destination websites were built to inspire. You rely on stunning photography, cinematic video, event calendars, and storytelling to make someone fall in love with your place. That still matters. Inspiration is the top of the funnel, and it always will be. But once a visitor knows where they are headed, they switch into planning mode, and that is where AI shifts from a novelty to a utility.

And only 20% of DMO leaders rate itinerary building as a high-value feature worth investing in.

We are ceding the highest-intent part of the planning journey (research, comparison, itinerary building) to third-party platforms that have no loyalty to your destination, likely no relationship with your local partners, and no incentive to distribute economic impact equitably across your community.

And AI is reshaping the competitive set between destinations while it does it. When a visitor searched on an LLM ideas on a vacation, they are discovering destinations they may not have thought about which means they can likely change their minds on where to go. AI is now an active participant in your comp set.

Your’e Not an OTA, and that is A-OK

I think a lot of the AI conversation in our industry accidentally creates anxiety about the wrong piece, and that’s the idea that we can control attribution by controlling where the booking occurs.

Becoming a direct booking engine is not, and has never been, the core job of a DMO. I am not going to sit here and tell you to start processing credit card transactions or competing with Expedia for commissions. Unless you want to activate hard mode.

The visitors agree (for now.) Expedia found 66% would not trust an AI assistant to buy or book anything on their behalf. PwC’s summer spending poll found one-third of respondents already use AI agents or bots to book parts of their trip. And remember the Amadeus gap from earlier: 65% want to book inside an assistant, 20% would let it spend the money. The comfort is arriving in stages, and every stage is mediated by the assistant.

Just because you are not processing the transaction does not mean you can afford to be absent from the booking journey. If AI assistants are building itineraries that include or exclude your local partners, you need to be feeding trusted, structured data into the places those assistants read. You exist to support your local hoteliers. Your goal is to make sure LLMs recommend your local hotel, your local restaurant, your local attraction operator, and then to prove that your influence drove that outcome.

Proving Value when the Click has Peaced Out

If your website traffic is dropping because AI is answering the visitor’s question before they ever reach your URL, how do you prove return on investment? How do you justify your budget? How do you show attribution when the digital funnel you were handed in 2006 is dying?

You can’t walk into a board meeting, point to a Google Analytics dashboard showing a 30% decline in organic sessions, and expect that to go well, right. And you certainly can’t rely on reports full of impressions and reach metrics that have always been, let’s be real, somewhat disconnected from economic reality.

Two layers, then. The AI layer, and the ground-truth layer.

The AI layer costs nothing and most teams haven’t opened it

Start with Google Search Console. Google folds the impressions and clicks you earn inside AI Overviews and AI Mode into the regular Search performance report, blended in with the blue links and unlabeled, so the click you got from an AI Overview has been sitting in your numbers. What was missing was any way to isolate it. The new generative AI performance report, rolled out to every site worldwide as of August 31, is that view: it shows how often your pages appeared inside AI Overviews and AI Mode, by page, country, device and date. Today it reports impressions only, no clicks and no position, and Google says more metrics may follow.

Google Analytics is the other half. The AI Assistant channel added to the default channel group back in May shows sessions arriving from ChatGPT, Gemini, Copilot, Perplexity and the rest, and it excludes AI Overviews and AI Mode, which Google files under Organic Search.

So, read the two together: Search Console for how often Google’s own AI features showed you, Analytics for who arrived from everyone else’s assistant. An analyst who reads the AI Assistant channel as “all AI traffic” is undercounting the biggest one. Beyond Google, Microsoft Clarity now reports page citations and share of authority across AI answers, and Bing Webmaster Tools has an AI Performance view for Copilot. Have your team work through gathering this.

A prompt to hand your team

Paste this into Claude or ChatGPT. If your assistant can drive a browser (Claude in Chrome, or ChatGPT’s agent mode), it can do most of the clicking while you watch. If it can’t, it will walk you through one screen at a time. Log in yourself and never paste a password into the chat.

****Paste this into Claude or ChatGPT*****


You are helping a destination marketing organization set up free measurement of its AI visibility. Our website is [paste your domain]. We use Google Search Console, Google Analytics 4, and Google Tag Manager [delete any we do not use]. Work through the five steps below in order. If you can control my browser, do the clicking, narrate what you are doing, and stop and ask me before you change any setting or install any tag. If you cannot, give me exact click-by-click instructions one screen at a time and wait for me to confirm before moving on.
Step 1, Google Search Console. Confirm our property is verified. Open the Performance section and find the Generative AI (Search) report. Show me impressions from AI Overviews and AI Mode for the last three months by page, by country, and by device, and export the table. Remind me that clicks from those features are already counted inside the regular Search performance report and are not labeled there.
Step 2, Google Analytics 4. Open Reports, then Acquisition, then Traffic acquisition, and confirm a channel called AI Assistant appears in the default channel group. Build a comparison of AI Assistant against Organic Search for the last 90 days: sessions, engagement rate, average engagement time, and key events. If our property is older than May 2026 and the channel is thin, help me create a custom channel group with a rule for sessions whose source matches chatgpt.com, gemini.google.com, copilot.microsoft.com, perplexity.ai, or claude.ai, so I can see history. Remind me that this channel excludes Google's AI Overviews and AI Mode, which sit inside Organic Search.
Step 3, Microsoft Clarity. Help me create a Clarity project for our domain and install its tag, through Google Tag Manager if we use it. Then open the AI citations view and show me page citations, AI referral traffic, and share of authority.
Step 4, Bing Webmaster Tools. Help me verify our site, importing from Search Console if that option is offered, and open the AI Performance report for Copilot citations.
Step 5, the monthly habit. Draft a one-page monthly template with four lines: AI impressions from Search Console, AI Assistant sessions from GA4, citations and share of authority from Clarity, and Copilot citations from Bing, each with a one-line note on what that tool does not count. Then draft the six questions I should send our visitor-data vendor: where the location panel comes from and what share of devices it represents, whether it is opt-in, their visitor definition, refresh cadence and lag, how they normalize for panel drift, and the margin of error on the figures in our dashboard.

The ground-truth layer is where your board lives.

Double down on proving real-world EI: visitor flows, foot traffic, and spend inside your destination. There are several vendors and partners in this space (Zartico, Datafy, Arrivalist, Placer.ai, Sojern and Tourism Economics among them), and they all measure this, in different ways. The goal is the whole picture: what geolocation, spending and lodging together can tell you about a visitor trend.

One thing that truly defines which partner makes sense for you is to ask for their methodology. Every one of these platforms works from a sample. Foot traffic panels see a slice of the phones that were there (the one peer-reviewed study puts a major U.S. panel at 7.5% of the population), card data misses cash, and every vendor models the rest. That’s fine as long as they can explain where the data comes from, how they define a visitor, and how they fill the gaps. The one that explains it clearly is the one your board can trust.

Five states now ban the sale of precise geolocation data outright (Connecticut’s ban takes effect October 1), and Delaware’s governor signed a bill on September 2 that puts contracting and due diligence duties on businesses that buy personal data, as well as the ones that sell it. Every location panel in this category is about to get smaller and more expensive, which makes that methodology conversation less optional every quarter.

Instead of walking into your board meeting and saying, “We drove 100,000 clicks to our website this quarter,” you can say:

“Our structured data and AI visibility work ensured our destination was recommended by AI assistants for the questions that matter, and our in-market dashboard shows a measurable increase in foot traffic and visitor spending in our core downtown zone over the same period.”

(Insert your own number, and make it one your vendor can defend.) These work in a symbiotic relationship. That is a narrative your board understands. That is a narrative your hotel partners respect. That is how you protect your budget when the funnel is invisible.

Two examples of those winning:

• Ljubljana Tourism won the Tourism category for LUV Fest, a festival the DMO designed and produced in-house to attack the fact that almost half of its overnight stays fell in four summer months, reaching 120-plus partners and 220-plus events this year.

• Visit Austin won Creative Design for a website rebuilt around AI-era discovery, which meant retiring 97 microsites, consolidating 500-plus pages, and removing 600-plus low-performing blog posts.

Talent Velocity, The Real Bottleneck and Priority

“This all makes sense, but I have a team of six people, a budget that barely covers our paid media, and I cannot hire a Chief AI Officer.”

You are right. And you don’t need all of this.

These reports confirm what you already feel in your bones. When DMO leaders were asked what stands in the way of AI adoption, the ranking was staff bandwidth first, technical expertise second, and budget third, a ranking Mindtrip’s Michelle Denogean confirmed to PhocusWire in August. This is a people problem before it is a budget line item problem.

And it is not unique to destinations. ManpowerGroup’s February survey of 39,063 employers found 69% of U.S. employers report difficulty filling roles (72% worldwide), with AI skills the single hardest category to find for the first time.

The people you already have are further along than you think. Gallup’s April read found half of employed U.S. adults now use AI in their role at least a few times a year, and 13% daily. And Gallup found employees whose manager actively supports the team’s AI use are more likely to say their workplace culture improved this year (31% versus 21%), while half of the CHROs in the same study are not confident in their managers’ ability to guide employees on using AI at work. Grant Thornton found the same org looks completely different depending on who you ask: 34% of CIOs and CTOs say the workforce is ready to adopt AI, against 7% of COOs.

Your team is already adapting. They are drafting social posts, summarizing meeting notes, and brainstorming campaigns with the assistants right now, organically, without centralized strategy, guidance, or training. The adaptation is happening. The velocity is not. What you need is not a new hire. What you need is talent velocity, the speed at which your existing team can learn, adapt, and apply new tools to generate measurable value. Without deliberate investment in upskilling, we fall into pilot purgatory: many small experiments competing for attention without ever achieving scalable impact. You need an adaptable team.

5 Plays without Breaking the Budget

You do not need to reinvent the visitor experience overnight. You do not need a five-figure tech overhaul. You need a focused playbook that builds a durable foundation for the AI era.

1. AUDIT YOUR ANSWER ENGINE OPTIMIZATION

SEO is no longer enough. AI assistants need structured, trustworthy data to recommend your destination. Conduct a rigorous audit of every digital asset you own and every listing your partners maintain. Google Business Profile, Yelp, TripAdvisor, and Apple Maps. Implement schema.org markup for your events, attractions, and businesses. Then open your robots.txt and your CDN settings and decide which AI crawlers you let in, because as of today Cloudflare may be deciding for you. If an assistant cannot read, understand, and trust your data, you do not exist to the visitor.

I made a free download for DMO teams here for you to use, to do this in-house. I recommend 3-4 staff, a Friday afternoon, and lots of snacks.

Download this AI Audit Kit

2. ACCELERATE TALENT VELOCITY OVER SOFTWARE SPENDING

You don’t need to be buying five AI tools nobody on your team knows how to use. Dedicate a meaningful slice of your budget to training the people you already have. Brand USA’s Janette Roush teaches a prompting framework called CRIT (Context, Role, Interview, Task) that gets a marketing team past the blank chat window in an afternoon. When your team can summarize community feedback, draft partner emails, read CoStar and KeyData trends faster, you multiply capacity without adding headcount. Gallup says the manager who visibly supports the team’s AI use is the difference. Be that manager.

3. RE-ARCHITECT YOUR MEASUREMENT FRAMEWORKS

If AI is intercepting discovery, reporting on web sessions and pageviews is like reading a thermometer to diagnose a broken bone. Turn on the GA4 AI Assistant channel, the Search Console generative AI report, and Microsoft Clarity’s citation view, and know what each one excludes. Track how often your destination is named for your key questions across the assistants that matter, by name, because ChatGPT’s share is down to about 53% and falling.

Most importantly, integrate a destination operating system to measure visitor flow, geographic mobility, and spending, and tie it into your AI measurements. Close that gap in your own shop first. Give your board ground-truth numbers, and prove your digital strategy is driving physical visits even when discovery happens inside an assistant you cannot see.

4. BUILD UTILITY ON TOP OF THE VISUALS

Your website was built to inspire, and that still matters. But most travelers use AI to research things to do and build itineraries. Consider a simple AI itinerary builder on your site to keep the visitor in your ecosystem and turn inspiration into a plan that features your local partners. And take Visit Austin’s lesson seriously: they retired 97 microsites and removed 600-plus posts before they added anything. The assistants reward a clean, answerable site over a sprawling one.

5. INFLUENCE THE AGENTIC BOOKING JOURNEY

Do not try to become a booking engine. Google’s AI Mode launched hotel booking with ten partners and none of them is a destination, which tells you where your lever is: upstream, at the moment the assistant decides which destination to name and which partners to describe. Make sure your destination’s structured data and your partners’ inventory information are accessible to the assistants, and work with the AI travel platforms that are willing to ingest official destination data. The window to establish your destination as a trusted source is open now. You win when your partners win, and you prove it with ground-truth impact.

The two-year gap is real, but it is not a life sentence. Visitors have already made AI a fundamental part of the journey. The organizations that adapt now have an opportunity to close the gap, strengthen their role as economic advocates, and help shape what comes next.

Start with your data. Focus on talent velocity. Measure ground-truth impact. Build for utility. And influence LLMs that are increasingly making decisions on behalf of your future visitors.

What is your biggest hurdle in implementing AI or measuring real-world impact at your destination? Hit reply and let me know. I read every single one.