GEO in Japan: measure the Japanese answers first

Generative engine optimization (GEO) is the work of getting a brand mentioned, recommended or cited in the answers that AI assistants such as ChatGPT, Gemini and Claude generate. In Japan the same work is usually called AIO, a term that in English more often refers to Google AI Overviews.

Ranking well in search does not guarantee that an AI assistant will recommend you. Assistants combine information in their own way, so search rankings and AI mentions have to be measured separately.

What an assistant recommends can depend on the language of the question. With web search on, a question in Japanese can bring up Japanese pages and the brands those pages discuss. A brand that is well known in English-language sources is not necessarily the one recommended in a Japanese answer.

Japanese answers may also write a brand name in katakana rather than in Latin letters, so a mention can be easy to miss if you only search for the spelling you use at home.

AIOPulse measures the Japanese side only. It does not measure answers in English, and it does not compare the two.

There is no setting that makes an AI recommend a brand, and nobody can promise a result. The practices most often suggested are the ones that also help in search, applied to Japanese sources:

• Publish clear, specific information in Japanese about what you sell and who it is for, including first-hand data where you have it.
• Add structured data (schema.org JSON-LD) so that the facts on your pages can be read by machines.
• Answer the questions customers actually ask, in a plain question-and-answer form.
• Earn mentions in Japanese media, reviews and comparison sites, the kind of sources an AI may find when it searches.
• Keep your brand name, and the way it is written in Japanese, consistent across your own pages.

Whether any of this works for your brand is something to measure, not assume.

1. 01

   Pick the questions by job, not by industry

   Customers ask an AI about the problem they want solved, not about an industry. Choose keywords that describe the job your product does in Japanese, and expect competitors from outside your category.

2. 02

   Record which brands appear, every day

   AI answers change from day to day. Ask the same questions daily and record whether your brand and your competitors are mentioned, so a single good or bad day does not mislead you.

3. 03

   Compare before and after a change

   When you publish new Japanese content or get press coverage, compare mention share before and after with the same question set. Without a fixed baseline, a change in the questions looks like a result.

• Are the questions asked in Japanese, and can you read them?
• Do the questions leave your brand name out? A question that names you inflates the result.
• Which models are asked, through the API or the app, and with web search on or off?
• Can you read the original answers behind each number?
• Is the share calculated among the tracked brands' mentions, or as a portion of all answers? These are different numbers (mention share and mention rate).
• Does the tool count your brand when an answer writes it in katakana or another spelling?
• When the questions or models change, is the date recorded, so trends stay comparable?

AIOPulse's answers to these questions are on the home page and in the glossary.