Who Represents a Destination in AI Travel Discovery?

Punta Cana is one of the Dominican Republic's most established resort destinations. Yet in seven Source Tracker scans associated with Punta Cana, the country's official tourism website was not retained once in Claude's or Gemini's source records.

During the same study, that website was retained in every scan associated with La Romana and Miches, across all four AI systems.

The publisher was the same. The observed source environment was not.

Cyprus revealed another version of the same contrast. Its national tourism website appeared in all 20 ChatGPT scans associated with our Cypriot destinations, but only one of 20 Perplexity scans.

I'm from the Dominican Republic. My co-founder, Alexandros Zisimidis, is from Cyprus. We chose these markets because we know the places and many of the hospitality businesses behind the names. But familiarity with a market does not tell us which information sources appear when AI systems answer travel questions.

If I marketed a hotel in Punta Cana, I would not read these results and immediately ask the Ministry of Tourism to change its website. I'd ask which sources appear around the questions my prospective guests are asking, what those sources say about my property, and which gaps I can actually do something about.

That's the commercial question behind this research. Having accurate information online is not the same as knowing where that information is represented in AI travel discovery.

What did we study?

Our first Anana Research essay examined 29,320 retained source-domain observations across 51 destinations. Here, we focused on nine: La Romana, Miches, Puerto Plata, Punta Cana, Samaná and Santo Domingo in the Dominican Republic; Ayia Napa, Limassol and Paphos in Cyprus. [1]

Anana's public Source Tracker sends 20 travel-discovery questions per destination to ChatGPT, Claude, Gemini and Perplexity. A domain is retained in a scan when it appears in source evidence returned for at least two successful responses. The systems expose different types of source evidence, so we analyze each separately.

Half the questions name the destination. The other half are open travel questions repeated across destinations. A source retained in a scan associated with Paphos is therefore not necessarily a page about Paphos.

The study covers 40 destination scans per engine in the Dominican Republic and 20 per engine in Cyprus, collected from July 27 to September 14, 2026. These are observations of source presence under a defined protocol, not measurements of recommendation influence or traveler behavior.

Does an official tourism website consistently appear?

Both countries maintain national tourism portals: Go Dominican Republic and Visit Cyprus. Yet their presence in the retained records was far from uniform. [2, 3]

Source: Anana Source Tracker, historical research cohort. Dominican counts include es.godominicanrepublic.com. The higher-collection subset includes 11 Cyprus scans with at least 72 of 80 total responses successful across the four systems. These are within-engine observations, not comparable measures of engine preference. [1]

In ChatGPT's records, both national tourism sites were almost permanent fixtures. Stored examples included destination-specific pages about Miches, Cape Greco and Kourion.

The other collection mechanisms told a different story. The Dominican domain met the retention threshold in at least half the scans on each. The Cypriot portal met it in only four Gemini scans, one Perplexity scan and none of the Claude scans.

Cyprus had less complete collection. We therefore repeated the check on its 11 better-collected scans and on the dates shared with the Dominican sample. The broad pattern remained.

This is not a scorecard for either ministry. Nor does a missing retained record mean the site never appeared in a response. Official status alone does not guarantee that a website will be part of the source set a hotel team observes.

For marketers, that changes the starting assumption. A national tourism campaign or a well-maintained destination portal may provide important information. Neither tells you, by itself, which pages your prospective guests' AI-assisted research will surface.

Why does the destination itself matter?

The contrast becomes even more useful when we stop comparing countries and look inside one of them.

Source: Anana Source Tracker, historical cohort. All but one of the 40 Dominican destination scans met the higher-collection threshold. [1]

La Romana and Miches retained the national tourism domain in every scan on all four systems. Punta Cana retained it in six of seven ChatGPT and Perplexity scans, but none of the seven Claude or Gemini scans. Puerto Plata also showed an uneven pattern.

We cannot establish why. It may be tempting to point to destination size or competing publishers, but the records do not test those explanations.

What they do challenge is the idea that a hotel group can infer the information environment of one property from another, even within the same country.

A hotel in Miches and a hotel in Punta Cana may need different source investigations, even if they share a national tourism authority. A portfolio-wide content strategy does not replace destination-level evidence.

Which publishers can a hotel actually work with?

National tourism websites are only one part of what appeared. The more useful examples show several ways relevant information can be published, and different actions a hotel might consider.

Regional tourism organizations. The Pafos Regional Board of Tourism appeared in all seven ChatGPT scans associated with Paphos. The Limassol Tourism Development and Promotion Company appeared in all seven associated with Limassol. A hotel might reasonably inspect its destination coverage on these sites before deciding whether a correction or a partnership is worthwhile. [4, 5]

Commercial destination publishers. ChooseYourCyprus combines destination content with accommodation and travel-booking options. Its domain was retained in all seven Perplexity scans associated with Paphos and 12 of 20 Cyprus scans overall, but none of the 20 ChatGPT scans. The Limassol tourism company's website also credits ChooseYourCyprus as its technology provider. That published relationship complicates any simple division between official and commercial information. It does not show that one website displaced another. [5, 6]

For a Paphos hotel, the takeaway is not to chase ChooseYourCyprus merely because it appeared repeatedly. It is to inspect the relevant pages. Is the property listed? Are its facilities described accurately? Does the page answer the questions the hotel wants to be considered for? If not, is there a reasonable editorial or partnership route to improve it? Those are concrete decisions a team can evaluate.

Hotels as destination publishers. Club Med's domain was retained in all six scans associated with Miches on every engine. The representative ChatGPT URL pointed to the Club Med Michès Playa Esmeralda property; Perplexity's pointed to a guide to things to do in Miches. The same hotel business appeared through both property information and destination content. [7]

That suggests two distinct content questions for a hotel marketer: are our rooms and experiences explained clearly, and are we answering useful questions about the destination that surrounds them? The data doesn't prove the guide improved Club Med's recommendations. It shows an observable example of a hotel operator publishing both kinds of information.

Property websites also featured. Casa de Campo was retained in all seven ChatGPT scans associated with La Romana. In Paphos, Almyra and Annabelle appeared in all seven. These are property-specific websites, not necessarily independently owned hotels.

The point is not that every hotel should create a destination blog. It's that information about a hotel can live on its own pages, regional tourism sites, commercial guides and other publishers. A useful strategy starts by finding the relevant gaps, not assuming that producing more content will solve them.

What would I do differently if I marketed a hotel in these markets?

I would start with the demand the hotel actually wants, not the broad question of whether the country appears in AI search.

A resort might want to win more family holidays, multi-generational stays or weddings. Guest inquiries can reveal which questions repeatedly stand between interest and a booking. That gives the team a sharper starting point than a generic list of travel prompts.

From there, the decision process is straightforward:

1. Measure the hotel, not only the destination. Run relevant traveler questions and record how the property and its competitors appear. Establish a baseline before changing anything.
2. Inspect the publishers that matter to those questions. Look at relevant source pages, not just recurring domain names. Separate information the hotel controls from pages owned by a tourism authority, booking partner or commercial publisher.
3. Choose one specific improvement. Correct an inaccurate description, clarify an important amenity on the hotel's website, or work with a relevant partner to improve factual coverage. Don't assume the most frequently observed source is automatically the best one to target.
4. Repeat the measurement. Did the hotel's description become more accurate? Did its recommendation frequency change consistently? Where tracking allows, follow the results through qualified inquiries and bookings, keeping direct and intermediary routes distinct.

This is where Anana's AI Source Intelligence and hotel-specific AI Presence workflow in Visibility Lab complement one another. The Source Tracker provides destination-associated observations. AI Presence measures how a particular hotel appears across defined questions and competitors, giving teams a way to test changes rather than assume their impact.

The Source Tracker findings do not tell us that changing a listing will raise visibility. They help a hotel decide which information to investigate before spending time and money on an experiment. That is a meaningful distinction for teams responsible for both marketing budgets and commercial results.

The next question for hotel marketers

The Dominican Republic and Cyprus are two case studies, not proxies for the Caribbean and Mediterranean. They reveal differences in retained source presence, not a ranking of countries, tourism authorities or AI providers.

But they do challenge an assumption that matters in everyday hotel marketing: publishing accurate information is not the same as knowing how that information appears around the questions prospective guests ask.

Hotels have spent years managing their representation across direct websites, OTAs and distribution partners. As travel discovery and booking evolve, they have another information environment to understand. The advantage comes from asking a better question.

Not simply, "Is our hotel online?"

But "Where is information about our hotel being represented, what does it say, and which change should we test next?"

Methodology and limitations

Collection. Anana's Source Tracker queried ChatGPT, Claude, Gemini and Perplexity via APIs with 20 English-language prompts per destination: one destination-named and one destination-agnostic prompt for each of ten traveler personas. A domain was retained if it appeared in source evidence from at least two successful responses in a destination-engine scan. The export aggregates by domain and saves one representative URL rather than full historical responses; Gemini does not provide comparable page URLs.

Sample. This case study uses the approved historical cohort from the first Anana Research essay: six Dominican destinations (40 scans per engine) and three Cypriot destinations (20 per engine), observed July 27 to September 14, 2026. Thirty-nine Dominican and 11 Cypriot destination snapshots met a threshold of at least 72 of 80 successful responses across all engines. Cyprus results were checked on that higher-collection subset and on shared scan dates. The threshold does not guarantee equal collection completeness for each engine.

Interpretation. ChatGPT records inline citations; Claude records retrieved search results; Perplexity combines search results and citations; Gemini records grounding domains. Cross-engine counts are not directly comparable measures of citation use. The destination-named and generic prompts are blended, so retained domains cannot be assigned to a specific question. A domain may appear once without meeting the retention threshold. Operator and page checks describe websites as seen in October 2026, not necessarily at collection time. This descriptive study does not establish causal influence on recommendations, traveler behavior or bookings, and the nine destinations are not a representative national or regional sample.

Frequently asked questions

Does a missing retained website mean the AI system never used it?

No. A domain had to appear in source evidence from at least two successful responses to enter this dataset. Some single appearances are not retained, and source-extraction mechanisms differ by engine.

Should a hotel prioritize every site that appears frequently?

No. Prioritize relevant pages that address traveler needs the hotel wants to serve, then assess whether there is an accurate, appropriate change the hotel or publisher can make. Frequency is a starting signal, not a guaranteed optimization lever.

How would a commercial team know whether an improvement worked?

Repeat comparable hotel-specific AI Presence measurements, examining accuracy, recommendation frequency and competitive context. Where available, assess referral visits, qualified inquiries and bookings separately rather than attributing every change to an edit.

References

1. Anana proprietary research. Source Tracker July-September 2026 historical case-study cohort, using the same measurement and exclusions described in Which Sources Keep Appearing in AI Travel Discovery?. The public Source Tracker is a live tool and may differ from this frozen historical study.
2. Dominican Republic Ministry of Tourism. Go Dominican Republic. Official tourism portal; accessed October 12, 2026.
3. Republic of Cyprus, Deputy Ministry of Tourism. Visit Cyprus. Official tourism portal; accessed October 12, 2026.
4. Pafos Regional Board of Tourism. Visit Pafos. Accessed October 12, 2026.
5. Limassol Tourism Development and Promotion Company. Limassol Tourism. Website credits ChooseYourCyprus as technology provider; accessed October 12, 2026.
6. ChooseYourCyprus. Destination information and travel-booking portal; accessed October 12, 2026.
7. Club Med. Things to do in Miches, Dominican Republic. Historical URL example; present-day site checked October 12, 2026.

Research observations reflect July-September 2026 scans. Contemporary website descriptions were checked October 12, 2026. Anana's linked product and Knowledge Center pages provide definitions and context, not independent validation of the findings.