Kamogawa AI Search Lab — Kyoto

Topic

GEO / LLMO

Testing the field called generative engine optimization through measurement.

3 pieces

Pieces on this topic

Report #000 2026.08.02 Monthly Report #000: How Generative AI Answers About Kyoto Experiences (July 2026) We asked ChatGPT for Kyoto tea ceremony, kimono, and zen meditation recommendations, 20 times each in four languages, and logged which sources it cited (12 cells, 240 runs). Which sources get cited flips entirely depending on the query language and category. → Feature 2026.08.11 Measuring AI Visibility Through the API Is "a Completely Wrong Approach" A Kyoto kimono-rental sub-study: ChatGPT's citation sources were completely different between the web version and the API. The affiliate site that took 90% of web-version citations appeared zero times out of 20 API runs. Independent tests from Surfer, Luxeo Team, and the University of Hamburg back this up. → Feature 2026.07.14 It had great Google reviews. The AI said it didn't exist. Ten cultural-experience businesses in Kyoto, queried in Japanese, English, Traditional and Simplified Chinese. A shop with a strong Google rating and plenty of reviews was told 'not found' by AI and described as a different business. The difference wasn't size or review count. →

Topic Index

Other themes

4pieces AI Search The process by which generative AI searches for, cites and summarizes information. → 2pieces Entity Whether AI recognizes a business as one distinct entity, or confuses it with another. → 2pieces Multilingual The same question in four languages. Change the language and the sources change too. → 2pieces Tea ceremony / Kimono / Zazen The three categories under continuous observation, each with a different level of commerciality. → 1pieces Methodology What does it actually take to say a business is "visible to AI"? → 1pieces Noise floor Answers vary even under identical conditions. We measure that variation first. → 1pieces Independent shops & small companies What can businesses without an advertising budget actually do? →