CASE STUDY
An AEO company
was not showing up in AI search
We made our own site the subject, because we have no client case to show. The 40 target questions and the measurement method go out first, and the weekly results across eight weeks go in as they are — including the ones that got worse.
What gets measured, and how
The measurement conditions go out before the results, because a figure with no conditions cannot be verified — we apply to ourselves the standard we demanded of other companies.
- Target questions
- 40 (10 concept · 10 agency search · 8 brand-named · 7 regional · 5 data)
- Answer engines
- 7 (ChatGPT · Google AI Overviews · Perplexity · Claude · Naver AI Search · Copilot · Gemini)
- Observation unit
- 40 × 7 = 280 cells
- Cadence
- Same weekday and time of day each week. Logged-out sessions, question wording fixed
- Citation rate
- (cells where cited) ÷ 280 × 100 — the denominator of 280 does not change
- Recorded per cell
- Whether cited · position in the source list · mention accuracy · competitors mentioned alongside · anything notable
Publication starts from round zero, once the baseline measurement is complete.
No round goes out before all 280 cells have been measured. Writing "improved" with no starting point would drain the meaning from every number that followed.
This series is a record produced by Navirang Co., Ltd., a company that sells AEO services, with itself as the subject. The sample is one company, it is not a controlled experiment, and there is no guarantee that what is observed here reproduces at another company. The measurement conditions and the raw data are published alongside, so please check and judge for yourself.
What is AI saying about your brand right now?
Our free audit finds out. We ask real questions in the major answer engines and report what was cited and what was said.
We reply within one business day.