Comparing how AI recommends competing destinations
Which destinations AI assistants recommend when travelers haven't chosen yet, and which websites they draw on.
Before a traveler names a destination, AI assistants answer broad questions such as the best food cities in a country or the best winter getaways. This guide shows, from TakeScout's repeated checks, which tracked destinations those answers recommend, how concentrated they are, and which websites the one assistant that shows its sources relies on.
Is AI recommending competing destinations instead of ours?
When travelers ask broad questions such as the best food cities in a country, AI assistants recommend a short list of cities again and again. In our data, five cities took 42% of those recommendations, and 30 of 60 tracked cities were never recommended on them.
Canada: of 210 recommendations of a tracked city, Québec City took 55, Montreal took 31, and Vancouver took 25; 23 of the 37 tracked cities were never recommended.
United States: of 203 recommendations of a tracked city, New York City took 38, New Orleans took 25, and San Francisco took 21; 7 of the 23 tracked cities were never recommended.
This measure currently covers cities; state, provincial and national destinations aren't included yet.
| Most-recommended national question | Cities recommended most (answers) |
|---|---|
| Best Foodie Cities in Canada | Montreal (12 of 12), Toronto (12 of 12), Vancouver (11 of 12), and Québec City (9 of 12) |
| Best Foodie Cities in the United States | Chicago (12 of 12), New Orleans (12 of 12), New York City (12 of 12), and Los Angeles (11 of 12) |
Basis: 2,940 answers to regional and national questions about 60 cities (each answer counted for each city it could recommend) · collected October 1 to October 14, 2026 · ChatGPT, Claude, Gemini, and Perplexity. Unbranded discovery questions that don't name the destination; sampled AI answers, not traveler behavior.
What it means for a destination marketer
Destination-specific questions show what AI says about a place once a traveler has named it; these broad questions show whether it is on the shortlist before that. A destination can do well on the first and be absent from the second.
Presence rate in the methodology · A worked example: Québec City · Browse tracked destinations
Which websites do AI assistants cite when recommending a destination?
A destination's own official tourism website is cited in about half of Perplexity's answers about that destination (50% across 121 destinations). Of the assistants we track, only Perplexity shows its sources; the sites it lists most often are Tripadvisor, Expedia, and OpenTable.
A destination's own official tourism website appeared in 7,247 of 14,610 Perplexity answers about that destination that listed sources (50%), across 121 destinations; for the median destination, 51% of its answers.
What counts as appearing: an answer counts when at least one of the sources Perplexity listed for it is on one of the destination's official tourism-board domains, subdomains included. Those domains were reviewed and approved by hand for every tracked destination in September 2026; three cities with no city tourism site have none and are left out, as are destinations with fewer than 20 answers listing sources.
Across all its answers, Perplexity listed 225,374 sources from 16,955 websites over the last 90 days; the five most frequent were Tripadvisor (14%), Expedia (2.8%), OpenTable (2.6%), Hotels.com (2.4%), and reddit.com (1.7%), together 23%.
Perplexity is the only tracked assistant that retrieves web sources for its answers. The others are queried without web search and return no sources, so these are not the sources behind their recommendations.
Basis: 14,610 Perplexity answers that listed sources · last 90 days to October 14, 2026 · Perplexity. Sampled AI answers, not traveler behavior.
What it means for a destination marketer
Which sites a retrieval-based assistant relies on is visible and can be tracked; for the assistants that answer from model knowledge, there is no source list to inspect.
Trusted sources across the network · A worked example: Québec City
How these figures are produced
Regional and national questions are asked in several wordings to every tracked AI assistant, without naming any destination; each answer is checked for the tracked cities it recommends or mentions. Only Perplexity retrieves web sources; its listed sources are counted per answer. Full TakeScout methodology.
Last updated October 14, 2026. Figures are recomputed from TakeScout's latest checks when the page loads. More questions: AI Visibility Questions.