Presence
Whether the place appears in AI answers at all, and in which answers.
Research · Places
How places are measured.
AIR APAC measures how AI answer engines represent a place when travellers ask about it. The method is called Asymmetric Visibility. This page says what is measured, how a measurement runs, when results publish, and why a place is never ranked against another place.
What is measured
Six things we read about every place.
The same questions are put to the same set of AI answer engines for every place in a cohort. What is recorded is the representation itself, not a satisfaction score. For each place we read:
Identity
Which version of the place AI engines present: which attractions, which framing, whose description.
Official assets
Which official tourism assets the engines reinforce, and which ones they ignore or replace.
Cited sources
Which sources the engines cite when they answer, and who therefore shapes the representation.
Language
How the representation changes across languages, where answers in a language can be reliably obtained.
Engine variation
How the representation varies from one AI engine to another.
Presence and displacement between places are recorded as exhibits with stated rules. Where a rule is keyword-based, its output is reported as keyword matching, never as semantic analysis. We do not publish an exhibit the instrument does not produce.
The measurement period
The questions are published before the answers are collected.
A measurement period opens when the question set, the engines queried and the scoring rules are frozen and published, and it closes when every result in the cohort has been published free to every place measured. The full rule, including what we will not sell while a period is open, is on the independence page.
Each new period is frozen and published from scratch. A period that inherits an unpublished method from the last one has not opened. Prior participants get no continuity of method and no influence over the next instrument.
Publication
Free to the place measured, gated on language review.
Every measured place receives its result free, in the same form and at the same time as every other place in its cohort, whether or not it has any relationship with AIR APAC.
Two publication rules are method, not preference. First, no per-place report is published for a place whose question set has not passed review by a native speaker of the query language and an expert reviewer. Second, where answers in a language cannot be reliably obtained, that is published as a finding about the representation, not corrected around: the systems answering travellers in that language are not currently auditable by this method, and the place is told so.
If a place finds an error in its result, we publish the correction.
Ranking
Places are never ranked against each other.
The company index and the place measure are different instruments with different units of analysis. The AI Readiness Index measures organisations and publishes company tiers. The place measure measures one place at a time and publishes no comparison between places, ever.
A place's report compares the place against its own prior measurement, where one exists, and against the published method. There is no public league table of cities and there will not be one. A ranked list of destinations from a small cohort identifies its members, reads as judgement to the governments concerned, and would depend on the sourcing depth of the cohort rather than the place.
A place result says nothing about the organisations located in it, and nothing about any individual. The three measures in this portfolio are separate instruments with separate units, and results are never converted from one to another.