RESEARCH · DATA
A figure that cannot be verified is only a claim. What we measured, how we measured it, and what we still cannot tell — all in one place.
THE PROTOCOL, IN THREE LINES
AVAILABLE NOW
The protocol we fix and hold to: unit of observation, the three metrics with their formulas, the limitations we have not solved, and the change log.
Research on AEO, GEO and citation in AI search, each summarised with its sample, conditions and the limitations the authors state. Plus a claims table collecting where they disagree.
STATUS
Status is written as it is. Measurement in progress is not presented as a result.
40 questions × 7 answer engines, measured weekly under fixed conditions. No round is published before the baseline completes. When it does, the question set, the aggregation method including the denominator, and the raw data are published together.
The top 10 web documents for two queries in each province, classified by hand. The report and the raw CSV are at public addresses; because it is a frozen snapshot it is not overwritten by re-measurement. The write-up exists in Korean only for now.
Logged from production for the crawlers that actually reach this site. ⚠️ Observation is by user-agent string with no reverse-DNS verification, so we do not describe them as verified bots — and a visit is not a citation.
The measured reports themselves — the provincial search composition, the crawler logs, the eight-week series — are written up in Korean for now, with the raw data at public addresses. Research (Korean) →
Send us a URL. We fix the question set and the conditions first, then hand back the record with the raw observations attached.
We reply within one business day.