Methodology
As observed on nomad.watch. When someone asks an AI assistant about a place to live or work, the assistant may fetch a Nomad Watch page live to answer. Those fetches are in our own Cloudflare logs. An assistant fetch is a successful page request by a user-initiated assistant agent: ChatGPT-User (ChatGPT, OpenAI), Claude-User (Claude, Anthropic), Perplexity-User (Perplexity) and smaller ones such as MistralAI-User and Gemini Deep Research, which count as "other assistants". Search indexers (OAI-SearchBot, PerplexityBot, Claude-SearchBot) and training crawlers (GPTBot, ClaudeBot) are not counted.
Each page is mapped to the city, country, programme or regime it is about. A share is an entity's part of all assistant fetches of its kind in the month, rounded to a whole percent. Topics group the page kinds (costs, visas, taxes, climate and best time, guides, communities, jobs, other) of pages about a specific place, programme or regime; tax treaty pages count under taxes.
Visibility bias. An assistant only fetches our page when someone asks and its search ranks us, so fetches measure demand times our visibility. New or improved pages can raise a share without any change in demand. Read the figures as relative interest on Nomad Watch, not as market size.
Thresholds. An entity needs at least 20 assistant fetches in the month before its share is shown; below that the page says "not enough data". Movers compare shares with the previous month against the median entity of the same kind, need the minimum in both months and leave out pages whose content changed. Only the top 10 of each list are public.
Referrals. AI-referred visitors are people who arrived from an assistant's answer. ChatGPT often sends no referrer or UTM tag, so its share of referred visitors is likely understated.
License this data
This page is a curated public view. The licensed AI travel demand dataset adds what it leaves out:
- counts, not just shares, for every entity;
- the long tail beyond the top 10, down to single cities and programmes;
- daily detail within each month;
- per-operator detail for every assistant, search indexer and training crawler;
- the full history since collection started;
- monthly files with a manifest and checksums.
Rights and quoting
© 2026 Nomad Watch. All rights reserved. Unlike the facts and figures elsewhere on Nomad Watch (CC BY 4.0), this dataset is not licensed for reuse. You may quote individual figures with attribution and a link to the edition.