How Can a DMO Make Its Official Website the Source AI Trusts?

A DMO's official website becomes the source AI trusts when it holds the most specific plain answer, which today means questions that name the place, narrow themes, itineraries and practical facts. Across 14,865 AI answers about 12 destinations Drifter AI monitors, official sites were cited in 39.5% of answers to questions naming the destination and 10.5% of the rest.

These are answers we recorded between September 15 and October 25, 2026, from OpenAI's models, Anthropic's Claude, Google's Gemini and Perplexity through their search-enabled versions, and from Google's AI Overviews for seven of the destinations. All 12 are tourism board websites of very different sizes. Every answer counted here showed at least one source, and a source counts as official when its link sits on the tourism board's own domain or one of its subdomains. Altogether the official sites were cited in 3,245 of the 14,865 answers.

Does it matter whether the traveler names your destination?

It matters more than anything else we measured. When the question named the destination, the official site was cited in 2,298 of 5,825 answers. When it didn't, as in "best destinations for brewery trips", the official site was cited in 947 of 9,040. That gap held for every one of the 11 destinations whose site was cited at all.

The mechanism is simple once you watch the engines search. A question that names your place sends the engine looking for your place, and your site is the obvious match for those words. A question that doesn't name you sends it looking for a category, and categories are where publishers, list makers and forums live. So the official site is strongest at the bottom of the funnel, after the traveler already has you in mind, and weakest at the moment they are choosing between you and somewhere else.

Who gets cited instead when the traveler doesn't name you?

Reddit, Wikipedia and travel magazines carry the discovery questions. In the 9,040 answers to questions that didn't name the destination, Reddit was cited in 2,059, Wikipedia in 1,482, Lonely Planet in 778, YouTube in 697, the National Park Service in 604 and Travel + Leisure in 591. All 12 official sites together were cited in 947.

Reddit was cited in more of these discovery answers than the official site for 9 of the 12 destinations. Flip to the named questions and the order flips too. There the official site was ahead of Reddit for 9 of 12, with 2,298 answers to Reddit's 1,369.

Travel magazines show the same split more sharply. Lonely Planet turned up in 778 discovery answers and 164 named ones, Travel + Leisure in 591 and 52, Condé Nast Traveler in 261 and 38. Engines use magazines to decide which places belong on a list, and then use your site to describe the place once it's on it. That makes earned media and the forums your visitors already post in part of your AI visibility, because that's where the shortlist gets written.

Which discovery questions can an official site still win?

Narrow themes are where official sites still get cited on discovery questions. For questions about a specific interest, like brewery trips, live music weekends or water activities, the official site was cited in 341 of 1,712 answers. For broad trip types, like nature trips or beach vacations, it was cited in 36 of 1,155. Both sets cover the same nine destinations.

The broad question is a contest every coastal or mountain place enters, and engines settle it with roundups. The narrow one is closer to what a tourism board actually publishes: a page about the breweries, a page about the music venues. Questions about a kind of traveler sat in between, cited in 86 of 493 answers to questions like outdoor weekends with young kids or trips for culture lovers.

Questions about value and reputation were the hardest. The official site was cited in 22 of 315 answers to questions like "best value destinations" and 16 of 300 to questions like "highest-rated cities for family trips". I think that's the engines behaving sensibly. A tourism board can't be the judge of its own rating, so for those questions the engines go to reviews and independent voices.

Which pages on an official site get cited?

Specific pages get cited far more than general ones. The 3,245 answers that cited an official site pointed to 1,859 different pages. We sorted them by what their web addresses say they are, so treat the groups as approximate:

• Things to do and themed guides, like a page on hiking trails or a food and drink guide: 1,422 answers, 442 pages, 10 destinations

• Itineraries, road trips and weekend guides: 699 answers, 172 pages, 8 destinations

• Practical planning pages, like weather, travel tips and getting around: 425 answers, 101 pages, 10 destinations

• Events and festivals: 418 answers, 180 pages, 9 destinations

• Homepages: 361 answers, 15 pages, 9 destinations

• Business listings: 208 answers, 86 pages, 9 destinations

The homepage looks respectable in that list until you notice it's one page doing the work of dozens. The pages that carry most citations each answer one smaller question: where to hike, what to do on a Saturday in May, how cold it gets in March. An AI answer is built from several of those smaller questions, so the site with a page for each one has more ways in.

Pages written for partners, press and meeting planners also got cited for six of the destinations. Media kits, member directories, meetings pages and trade-site PDFs all turned up in answers to travelers. Whatever is public on your site is part of the source material, whether or not you wrote it for travelers.

What does AI read on your site but leave out of the answer?

AI reads a lot more of an official site than it cites. For one destination, we compared the pages OpenAI's search retrieved with the pages its answer actually cited. Across 274 of its answers from September 29 to October 23, OpenAI read the official site's pages 1,202 times without citing them, and cited them 388 times.

The homepage was read 57 times and cited in none of those answers. Itinerary pages did best, cited 103 times out of 339 retrievals. Themed guides were cited 209 times out of 846, practical planning pages 25 out of 97, and event pages 18 out of 117.

My read is that the engine pulls in several pages and then cites the one whose text actually states the claim it wants to make. A homepage tells the engine it has found the right place. An itinerary says which neighborhood to walk on day two and where to eat after, in order, and that's a sentence an answer can lean on. This is one destination and one engine, so take it as a strong hint rather than a rule.

Why are old PDFs answering for some tourism boards?

PDFs on a website vendor's file host were cited in 436 answers, and they don't count as the tourism board's own site. They sit on assets.simpleviewinc.com, the file host of Simpleview, a website platform many DMOs use. In our data they came from 45 different accounts on that host, almost all of them tourism organizations by their account names, and turned up in answers for 10 of our 12 destinations. 298 of those 436 answers were to discovery questions, where official pages are scarce.

Most of it is useful material. Visitor guides and printed itineraries made up 54 of the 176 files and were cited in 209 answers. But 32 files carried a year of 2023 or earlier in their names and were cited in 63 answers, and 18 were business documents like requests for proposals, sponsor packages and meeting packets, cited in 56.

There are two problems here. When AI quotes your visitor guide from a vendor's domain, your own site gets no credit in any count of citations, including your own reports. And when it quotes an out-of-date guide, the traveler gets out-of-date restaurants and hours. A tourism board can fix both by publishing the guide's best sections as web pages on its own domain and retiring the PDFs it no longer stands behind.

Does the engine change the picture?

It changes the numbers a lot, partly because engines show very different numbers of sources. On questions naming the destination, the official site was cited in 909 of 1,501 OpenAI answers, 550 of 1,225 from Perplexity, 440 of 1,359 from Gemini, 284 of 1,176 from Claude and 115 of 564 from AI Overviews. On discovery questions it was 536 of 2,206 for OpenAI, 254 of 1,982 for Perplexity, 79 of 1,905 for Claude, 70 of 2,054 for Gemini and 8 of 893 for AI Overviews.

OpenAI averaged 27.0 sources per answer and Perplexity 14.6, while AI Overviews showed 8.4, Claude 5.8 and Gemini 4.7. An engine that lists 27 links has more room for a tourism board than one that lists five. So a single blended citation number for your site mostly tells you which engines were in the mix. It's worth reading each one separately, which is how we measure AI visibility.

What should a tourism board do this week?

Start with the pages engines are already reaching for, then fill the gaps where they can't find you:

1. Pull your site's 20 most visited pages and mark which ones answer one specific question. A homepage, a category landing page or a "plan your visit" hub doesn't count.

2. Pick three narrow themes your destination is good at, like breweries, a trail system or a music scene, and check that each has its own page that names the places, the season and the practical details in plain text.

3. Turn your best visitor guide itinerary into a web page on your own domain, with the stops in order.

4. List every PDF your site links to, including the ones on your platform's file host. Update or remove anything dated 2023 or earlier, and take down business documents that travelers don't need to see.

5. Ask three discovery questions you want to win in at least two engines, and note which magazines and Reddit threads they cite. Those are your earned media and community targets.

What can't this data tell you?

These are 12 destinations Drifter AI monitors, not a random sample of tourism boards, and one large destination contributes more answers than the smaller ones. The page groups come from web addresses, so a themed guide filed under a blog address counts as a blog post. The split between pages read and pages cited comes from one destination and one engine. A citation shows which source an engine chose to display. It doesn't prove which source shaped the words, and it says nothing about whether the traveler clicked.