How to measure AI visibility for tourism destinations
A practical framework for tracking whether AI assistants recommend a destination — and how often they recommend somewhere else instead.
AI visibility measurement for tourism destinations means tracking, on a recurring basis, whether ChatGPT, Claude, Gemini, and Perplexity recommend a destination when travelers ask real trip-planning questions. It is not one test: in TakeScout's latest checks, every assistant that answered named the same place first in only 11% of questions (809 of 7,437). This guide shows how to measure that, using live figures from 124 tracked destinations.
Last updated October 30, 2026, from TakeScout's latest data.
Key numbers
- 11% of questions where every assistant that answered named the same place first
- 87 of 21,934 unique businesses AI put in its top three are confirmed permanently closed and still recommended by 2+ assistants
- 52% of Perplexity's sourced answers about a city cite its official tourism site (median of 57 cities)
- 4% median presence rate of 60 cities in regional and national questions that don't name them
- 48% of first picks changed between a destination's last two checks (same question, same assistant; 39 destinations with two complete checks)
- 124 destinations tracked · 109,756 questions asked · 136,165 places identified
Brand visibility vs. recommendation visibility
These sound similar but answer different questions, and the difference matters for what a tourism board should actually measure.
Brand visibility asks whether AI knows a destination exists — can it describe the place, does it recognize the name, does it have basic facts right. Most "AI visibility" tools stop here.
Recommendation visibility asks something more specific: when a traveler hasn't decided where to go yet, does AI actually suggest this destination — and how often does it suggest somewhere else instead? A destination can have strong brand visibility (AI describes it accurately when asked directly) and weak recommendation visibility (AI never surfaces it unprompted, in the open-ended questions travelers are actually asking while still deciding).
The second question is the one that affects trip volume. A tourism board doesn't need AI to know facts about them — a Wikipedia page does that. It needs AI to put them on the shortlist before the traveler has chosen. That's the gap this framework is built to measure.
Why this is different from SEO
AI answers aren't a fixed ranking. The same question can return a different destination depending on the model, the exact wording, and the day you ask — which means a single test tells you almost nothing. A destination might be consistently recommended for "romantic weekend getaways" while being invisible for "family road trips," and that split is only visible if you're testing multiple question types across multiple models over time, not checking once and calling it done.
This is also why a single "AI visibility score" can be misleading on its own. A destination can have middling overall visibility and still dominate a specific category — food travel, accessible outdoor trips, budget weekends — and that category-level detail is usually more useful to a marketing team than one composite number, because it points at something specific to act on.
What we've observed so far
Each finding below comes from TakeScout's tracking, with the figure behind it and where to see it.
One check isn't enough
Between a destination's last two complete checks, the same assistant's first pick for the same question changed in 48% of cases (3,407 of 7,088 comparisons across 39 destinations, checks from August 27 to October 14, 2026). This figure covers 39 of the 124 tracked destinations: only those whose last two checks were both complete. Most September checks were incomplete after an assistant outage, so the sample is small and may not be typical; it widens as more checks complete. Measuring means repeating the same questions, in the same wording, with the same assistants, and comparing like with like. One check can't separate a real change from this background variation.
Models disagree
Every assistant that answered named the same place first in 11% of questions (809 of 7,437). Two assistants at a time, first picks matched in 24%–35% of questions. Part of this is randomness within each assistant: between two checks, the same assistant changed its own first pick in 48% of cases. Low agreement between assistants reflects both genuine differences and that background variability. A test run on one assistant describes that assistant, not AI in general.
Model agreement · One check isn't enough
Wording matters
We ask each question in three wordings. Two wordings of the same question, put to the same assistant, share on average 33% of the places they recommend, and 0.6% of wording pairs (173 of 29,537) return exactly the same places. Answers collected from October 1 to October 14, 2026. A check built on one phrasing measures that phrasing.
TakeScout methodology · Québec City on TakeScout
Accuracy is part of visibility
At least 87 of the 21,934 unique businesses AI assistants put in their top three (0.4%) are confirmed permanently closed and still recommended by two or more of them (October 2026 update). A recommendation that sends a traveler to a closed business is visibility that works against the destination, so accuracy belongs in the measurement.
Your official site may not be the source
Perplexity, the only tracked assistant that lists its sources, cites the median city's official tourism site in 52% of its sourced answers (57 cities, answers from the last 90 days). The share runs from 18% for New York City to 83% for Québec City. For states and provinces, the median is 51% (62 destinations).
State of AI Travel · Trusted sources
Presence outside your own questions
Asked regional and national questions that don't name them, the median tracked city has a presence rate of 4%, and 30 of 60 cities are at 0%. The leaders: Québec City 100%, New Orleans 92%, and New York City 92%. Rates as of October 30, 2026.
Presence rate · Québec City on TakeScout
What is AI answer engine optimization (AEO) for destination marketing?
AEO is the practice of making a destination's official information easy for AI systems to find, trust, and cite when answering traveler questions — the AI-era counterpart to SEO. For a destination, this isn't a single trick; in practice it's a combination of:
- Entity clarity — structured, factual information about the destination (attractions, neighborhoods, seasonal context) that AI systems can parse cleanly
- Third-party authority — coverage from travel publishers, review sites, and other independent sources, since AI systems weigh outside validation heavily, not just what the DMO says about itself
- Source and citation visibility — knowing which websites AI actually cites when it talks about a destination, so gaps in the destination's own authority become visible rather than guessed at
The goal isn't to "trick" AI into recommending a place — it's closing the gap between what a destination promotes about itself and what AI already associates with it, using real information rather than repetition.
How can a tourism board track how AI recommends their destination?
A credible first version of this can be built with three ingredients: a question library, a testing routine, and a scoring framework.
1. Build a traveler-question library. Real trip-planning questions, not generic destination-name searches — "best European destinations for families in October," "affordable destinations with great food," "where should I go hiking without a car." Segment by season, traveler type, trip length, and budget, since visibility often varies sharply by category even when the overall number looks stable.
2. Test across the models travelers actually use. At minimum, ChatGPT, Claude, Gemini, and Perplexity. Run the same questions on a regular schedule, with the same wording each time — one-off checks don't tell you whether what you're seeing is a real pattern or normal day-to-day variability.
3. Record more than "did we get mentioned." For every answer worth capturing:
- Was the destination recommended, or just mentioned in passing?
- How prominently — top pick, or buried in a longer list?
- Which competing destinations showed up in the same answer?
- What attributes did the AI associate with it (food, affordability, outdoors, luxury, family-friendly)?
- Were the details accurate — are recommended businesses still open?
- What sources or citations backed the answer — the destination's own site, a review platform, a travel publisher, a forum?
4. Turn it into a few real KPIs, not a single vanity score:
- Recommendation rate — the share of relevant questions where the destination gets recommended
- Category ownership — which specific traveler intents (not overall visibility) the destination actually leads on
- Source/citation share — which websites are doing the work of supporting AI's answer, and whether the destination's own site is one of them
- Change over time — the same metrics, tracked longitudinally, since a single snapshot can't distinguish a real shift from normal variability
How can a DMO establish an AI visibility baseline?
Ask a fixed set of traveler questions, in fixed wordings, to the same AI assistants, and record every answer once in full. That complete first check is the baseline every later check is compared with.
Each TakeScout destination check asks 20 traveler questions about attractions, day trips, experiences, hotels, and restaurants, each in three wordings, to four AI assistants: 240 answers per destination.
Cities are also measured on regional and national questions that don't name them: 48 to 60 answers per city, across 60 cities.
Two checks are compared only when the same assistants answered, the question wordings are the same, and each check got at least 75% of its expected answers. In the latest round, 124 of 124 destinations' checks met that bar.
Basis: 29,760 answers to destination questions, 124 destinations' latest checks · checks run September 28 to October 14, 2026 · ChatGPT, Claude, Gemini, and Perplexity. Sampled AI answers, not traveler behavior or conversation volume.
What it means for a destination marketer
Changing the questions, the wordings or the assistants starts a new baseline: a comparison across them would show the change in method, not a change in what AI recommends.
The most useful board-level summary compresses all of this into a sentence or two a director can act on, rather than a single score. The worked example below ends with one generated from Québec City's live data.
Worked example: Québec City
The guide's KPIs, applied to Québec City from its latest check (October 2, 2026).
- Recommendation and presence rate
- Québec City comes up in Canada-wide and regional questions that don't name it, 1st of 37 tracked Canadian cities: presence rate 100% (5 of 5 applicable questions; recommended in 15 of 15 wordings).
- Category ownership
- Recommended in every answer to “Best Romantic Weekend Destinations in Canada”, “Best Winter Destinations in Canada”, and “Three-Day Itinerary in Québec Province”.
- Recommended in 10 of 12 answers to “Two-Week Must-See Itinerary in Canada”.
- Recommended in 9 of 12 answers to “Best Foodie Cities in Canada”.
- Source/citation share
- Perplexity cited quebec-cite.com, Québec City's official site, in 51 of 60 sourced answers (85%) in the October check, about the same as in August (80%).
- Accuracy
- Two places still recommended by two or more assistants are flagged closed: Chez Muffy and Restaurant Initiale.
- Change over time
- 38% of first picks changed since its previous check (68 of 181 comparisons, same question and assistant).
Board-level summary, generated from this data: “Québec City came up in 5 of 5 regional and national AI travel questions this check, ranking 1st of 37 Canadian cities. Its official site backed 85% (51 of 60) of Perplexity's sourced answers about it. Two closed places are still recommended to travelers.”
How do you know what ChatGPT says about your city or region right now?
The free, manual version: open an incognito or temporary chat (every major assistant has one — it keeps the session from being shaped by prior history) and ask a real trip-planning question about the destination, exactly as a traveler would phrase it. Do this as a single, standalone question per session, not a follow-up in a longer conversation — anything asked earlier in the same session can quietly shape the answer that follows.
This works fine for a spot-check. Where it gets impractical is doing it consistently enough to matter: a handful of questions across several assistants, repeated on a schedule and tracked over months, to tell a real trend from normal noise. That's the gap a dedicated tracking approach — whether built in-house or through a monitoring tool — is actually solving.
What to look for in an AI visibility tool for travel and tourism
Not all tools in this space answer the same question, and it's worth being clear about which one you actually need before choosing one:
- Single-property tools audit one destination or property at a time — useful if the goal is fixing your own site's AI-readiness and getting concrete content/technical recommendations for that one place.
- Network or market-benchmarking tools compare a destination against peers at once — a composite score against a network average, category-by-category standing versus named or unnamed competitors, and how the whole landscape is shifting over time. This matters more when the real question isn't just "are we visible" but "how do we stand relative to the destinations travelers are actually choosing between."
TakeScout is built around the second approach: ongoing, cross-model observation across a network of tracked destinations, with benchmarking, category-level detail, and source-citation analysis, rather than a one-time audit of a single place. And in line with the recommendation-visibility distinction above, it doesn't just track whether AI mentions a destination — it tracks whether AI recommends it, which traveler questions it wins, which destinations get recommended instead, and how that position shifts over time. See it on any tracked destination's report — browse the destinations directory.
Questions tourism marketers ask
Is AI visibility the same as SEO?
No. SEO measures where a page ranks in search results; AI visibility measures whether an assistant recommends the destination in its answer, which shifts with the assistant, the wording and the day.
Ask one question two ways and the places differ: two wordings share 33% of their recommendations on average, and only 0.6% of wording pairs (173 of 29,537) match exactly, in answers collected from October 1 to October 14, 2026.
Assistants also disagree with each other: see Models disagree · TakeScout methodology
How can a tourism board get its destination recommended by ChatGPT?
Nobody can guarantee that ChatGPT will recommend a destination. What we can show is how ChatGPT (GPT-4o, which we query without web search) treats the cities we track, and which ones it recommends.
Across 735 ChatGPT answers to the regional and national questions that apply to the 60 cities we track, it recommended the city in 18% of them (130) and left it out of the other 82% (605); it never named a city only in passing. Answers collected from October 1 to October 14, 2026.
On the same questions, the share of answers recommending the city was 15% (111 of 735) for Claude, 12% (90 of 735) for Gemini, and 11% (82 of 735) for Perplexity.
The cities ChatGPT recommended most often: Québec City, in 15 of 15 answers (Best Foodie Cities in Canada, Best Romantic Weekend Destinations in Canada, Best Winter Destinations in Canada, Three-Day Itinerary in Québec Province, and Two-Week Must-See Itinerary in Canada); New York City, in 9 of 12 answers (Best Foodie Cities in the United States, Best Romantic Weekend Destinations in the United States, Best Winter Destinations in the United States, and Two-Week Must-See Itinerary in the United States); and Ottawa, in 9 of 12 answers (Best Foodie Cities in Canada, Best Winter Destinations in Canada, and Two-Week Must-See Itinerary in Canada).
Presence rate in the TakeScout methodology · Traveler intents · Québec City on TakeScout · New York City on TakeScout
What is AI visibility in travel marketing?
For a destination, AI visibility is how often AI assistants name it when travelers ask where to go or what to do, and whether they recommend it or only mention it. It is measured across traveler intents and across assistants (ChatGPT, Claude, Gemini, and Perplexity), because each one answers differently.
The spread is wide. Across the 60 cities we track, presence rates (the share of regional and national question wordings in which an assistant recommended the city, a mention counting half) range from 0% to 100%: Québec City is highest at 100%; 30 cities, including Abbotsford and Atlanta, are at 0%. Rates as of October 30, 2026.
Presence rate in the TakeScout methodology · Québec City on TakeScout · Browse every tracked destination
What's the difference between being mentioned and being recommended by AI?
A recommendation is a place an answer puts forward as a main or strong pick; a mention names it as a secondary option, an alternative or in passing.
In TakeScout's presence rate, a recommendation counts in full and a mention counts half.
For example, Montreal was recommended in 12 of the 15 regional and national question wordings that apply to it and only mentioned in 2, so its presence rate is (12 + 2 × 0.5) ÷ 15 = 87%. As of October 30, 2026.
Counted per answer, across all assistants, 413 of 2,940 answers recommended the city they were checked for and 23 only mentioned it. Answers collected from October 1 to October 14, 2026.
Presence rate in the TakeScout methodology · Montreal on TakeScout
Can you measure AI visibility across countries or languages?
Across countries, partly; across languages, not yet: every question we track is asked in English, and cross-market and language comparisons are planned, starting with French.
Today we track 37 cities in Canada and 23 cities in the United States. Their average presence rate is 18% in Canada and 31% in the United States, and 23 of the cities in Canada and 7 of the cities in the United States are at 0%. As of October 30, 2026.
This is not a like-for-like ranking of the countries: the city lists aren't matched (they differ in number and size) and each country's cities are scored on that country's own national and regional questions.
Browse destinations by country · Presence rate in the TakeScout methodology
Can hotels measure their AI visibility the same way?
Hotels appear in our destination checks as their own category, so the same signals can be read for them, but TakeScout tracks destinations, not individual hotels.
On the five hotel questions we ask about every destination (Best Boutique Hotels, Best Budget Hotels, Best Family Hotels, Best Luxury Hotels, and Best Romantic Hotels), all the assistants that answered named the same first pick in 8.0% of comparisons (149 of 1,858), against 18% for attraction questions (408 of 2,231) and 5.9% for restaurant questions (110 of 1,860).
Among hotel questions, agreement is highest for Best Luxury Hotels at 16% (58 of 372) and lowest for Best Budget Hotels at 1.9% (7 of 371).
Latest answer per question and assistant, collected from October 1 to October 14, 2026.
Best Luxury Hotels: what AI recommends · Model agreement across the network
How these numbers are produced: TakeScout methodology.