Does AI recommend
our company?

September 9 follow-up: named searches showed CHIERO in all three services.

We checked ChatGPT, Gemini and Perplexity in their consumer interfaces. When asked what CHIERO and Ko Nakata do, all three mentioned the company and cited its official sites. For a separate question seeking an AI-informed business adviser without naming CHIERO, none of the three answers listed CHIERO as a candidate.

ChatGPT described the current advisory, publishing, art and app activities. Gemini mixed in an outdated investment description; Perplexity mixed older business descriptions with current information. Visibility and accuracy are separate observations.

Two questions, one fresh temporary or incognito conversation per service and question: six accepted answers. We explicitly requested web search. Account and regional effects were not fully controlled. These conditions differ from the August API study, so this is not a measured improvement rate or evidence of a ranking. Exact Japanese questions, selected answer excerpts, citations and conditions

Recommendations appeared only in questions that named us.

We asked 20 questions three times through each of three search-enabled provider APIs: 180 attempts, 158 successful answers and 22 errors. The saved classifications contain 17 estimated recommendations, all within two questions that included the company or representative’s name. The other 18 questions produced 140 successful answers and no estimated recommendations.

This is a dated observation of our own company, not our current ranking or evidence of results for clients.

These categories overlap. A citation is not necessarily an endorsement; a recommendation is not a lead or a sale. Recommendation labels were estimated from wording and list placement, not supplied by the providers.

What the note.com numbers mean.

The saved source-domain lists included note.com in 42 of 158 successful answers, across 11 of the 20 questions. There were 128 URL occurrences, including duplicates within answers. This does not mean CHIERO’s own note articles were recommended 42 times.

Limits of the measurement.

• The questions were designed for our own study, not sampled to represent all users. One named question used an outdated investor description. See our current company profile for our present activities.
• Twenty-two failed attempts are excluded from the answer denominators.
• Eight changes were published together, so this record cannot isolate the effect of an individual change or establish causation.
• Provider APIs do not establish what every signed-in consumer interface displays.
• We separately observed one unbranded question in four consumer interfaces on the same date. None showed the target in the four recorded categories. Four observations do not prove universal absence.
• Visibility cannot establish inquiries, contracts or revenue.

A repeatable way to observe AI search.

1. Choose questions from actual customer needs; keep named and unnamed questions separate.
2. Record the date, service, interface, visible model, language, location, search setting and conversation context.
3. Keep the answer, source URLs and failures. Classify retrieval, mention, citation and recommendation separately.
4. Use search-platform reports to check page discovery and visibility. Measure visits and business outcomes independently.
5. Repeat under comparable conditions after recrawling. Treat changed question sets as different studies.

Google says its usual SEO fundamentals apply to AI features; no special AI files or schema are required. Eligibility does not guarantee inclusion. Google’s documentation

OpenAI documents separate controls for search crawling and model training. OpenAI’s crawler documentation

Records and sources.

• API totals and source-file verification information
• Consumer-interface observations
• Original journal entry (Japanese)
• Full Japanese article

The original API record matches the published source-file hash, checked September 9, 2026. This article describes our research, not a service promising search placement.