Methodology
Our rankings measure what AI assistants actually recommend when a visitor to Avoriaz asks them.
1. The questions
For each category, we ask ChatGPT the questions a holidaymaker would ask: a general question ("Where can I eat the best fondue in Avoriaz?"), a top 10 request and specific needs (with kids, on a budget, English-speaking instructors…).
2. Two languages, two rankings
Answers are collected from the ChatGPT web interface as users see it: in English from the United Kingdom for the English version of the site, in French from France for the French version. The two rankings are calculated separately, because the AI does not always recommend the same places depending on the language. Each question is asked several times, since AI answers vary from one time to the next.
3. The score
- AI visibility (75%): each recommendation earns points according to its position (no. 1 = 10 points, no. 10 = 1 point).
- Review trust (15%): the Google rating, smoothed with a Bayesian prior (25 reviews at 4.6★), so that 5★ from 9 reviews does not beat 4.9★ from 150.
- Website citations (10%): a bonus when the AI cites the business's own website as a source.
4. What we publish
Only the top 5 places get a rank: they stay stable when the questions are repeated, unlike the following ones. Other places recommended several times are listed with how often they are recommended. A place mentioned only once is not shown. Categories where the AI does not name enough places (snowmobiling, doctors…) are not ranked. Pages for specific needs are only created when the AI recommends different places than in the overall ranking.
5. Independence
No business can pay to appear in a ranking or to move up.