Diagnosing Why Your Company Does Not Appear in AI Search

Tomohiro Iida · Published July 29, 2026 · Updated August 6, 2026

An AI service returns only your competitors. Your company name appears, but the description of your service is wrong. You are listed among the sources, yet no enquiry follows. These are not the same problem.

To improve a situation where your company does not appear in AI search, separate the stage at which the process stops. Publishing more articles resolves some cases. In others the cause is crawling, inconsistent information, missing comparison material, or the paths inside the site.

This article diagnoses five stages: retrieval, candidacy, citation, answer, and action. For each stage it sets out what to check first, and what that stage cannot tell you.

It is written for staff at B2B companies who can change both the content and the technical settings of their own site.

The limits, stated up front: the criteria by which a generative AI service selects the pages it cites, and the way it summarises them, are not published. The five stages below are a way of separating causes within what public information allows. They are not a procedure that guarantees inclusion or citation in AI answers.

What this article covers

• Covered: separating the situation into retrieval, candidacy, citation, answer, and action.
• Covered: deciding, for each stage, what to check first and what that stage cannot determine.
• Covered: converting the diagnosis into an improvement specification with an owner and a deadline.
• Not covered: only the stage of not being used as a source — see the linked article.
• Not covered: only the comparison stage, where competitors are recommended instead — see the linked article.
• Not covered: only the retrieval and indexing stage, from a technical angle — see the linked article.
• Not covered: fixing answers in which the company information itself is wrong — see the linked article.
• Not covered: definitions of the terms AIO, GEO, and LLMO — see the linked article.

Seven reasons AI search does not cite you as a source

Why AI search recommends only your competitors

Why search engines and AI cannot find your site, and how to fix it

Why AI states your company information incorrectly, and how to correct it

What AIO, GEO, and LLMO mean

The five stages covered here

Work through them from the top. While an upper stage remains blocked, fixing a lower stage on its own is less likely to show up in the result.

Stage 1: Is the page being retrieved?

Start by checking whether search-engine and generative-AI crawlers can reach the public page and read its body text.

• Whether robots.txt blocks the crawler in question.
• Whether authentication, a WAF, or geographic restrictions prevent retrieval.
• Whether noindex is present.
• Whether canonical points to a different URL.
• Whether the body text is in the HTML before JavaScript runs.
• Whether the page can be reached from the sitemap and internal links.
• Whether 404s or redirect chains remain.
• Whether crawler visits appear in the server log.

If there is a problem here, adding more content will not make the page retrievable. At the same time, allowing a crawler does not guarantee citation or recommendation. What this stage secures is retrievability, nothing more.

Google does not publish a dedicated schema or a dedicated file for AI Overviews or AI Mode, and states that the foundation is ordinary crawling and indexing (Google official guide, checked August 6, 2026). For crawler identifiers and permission settings related to ChatGPT, consult the OpenAI documentation. For generative AI services other than Google, check each provider’s published information separately.

Sources: developers.google.com/search/docs/fundamentals/ai-optimization-guide and developers.openai.com/api/docs/bots.

Deciding whether to allow or block AI crawlers

Checking whether a JavaScript site’s body text reaches crawlers

How to investigate Crawled – currently not indexed（日本語）

Confirming AI crawler visits in the server log

Stage 2: Are you in the candidate set?

Next, check whether you publish information that allows you to be selected as a candidate for the questions users actually ask.

In the range we see in our own diagnoses, company sites tend to centre on what the company wants to explain. Users, by contrast, ask with concrete conditions.

• The cost of an AI agent PoC.
• RAG adoption in manufacturing.
• A company in Kyoto to consult about Web3 development.
• A company that can go from SEO analysis through to code changes.
• Support for generative AI pilots aimed at local government.
• Whether a small company can commission the work at all.

If no page answers those conditions, there are fewer points at which you can enter the candidate set.

Information to publish in order to be considered

• Collect the questions customers ask from your enquiry history.
• State the industries, problems, regions, company sizes, and budgets you serve.
• Write both what you handle and what you do not.
• Explain the conditions that determine cost.
• Publish case studies and deliverables.
• Increase the information needed to make a purchase decision, rather than company introduction.

Rather than mass-producing near-identical articles for each search term, answer the background of the question and the decision behind it.

Classifying and designing for search intent

Why AI search returns only your competitors

Stage 3: Are you a source that gets cited?

Even when you enter the candidate set, your URL may not be offered as a source.

Which sources are cited differs by generative AI service, and the selection criteria are not published. For that reason, no single factor such as backlinks or structured data can be identified as the direct cause. Within what can be checked, put the following in order.

• The claim and its evidence sit on the same page.
• The method, scope, and reference date of any survey are written down.
• The author and the operating company are clear.
• There is primary information, not a summary of other companies’ articles.
• Fees, track record, and specifications match the source of truth.
• There is an update date and a change history.
• Proper nouns and service names are written consistently.

Rather than mechanically shortening text in the hope of being quoted, give priority to an information design in which a quotation does not create a misunderstanding.

Seven reasons AI search does not cite you

Being cited and being reflected in the answer are different things

Stage 4: Is the answer accurate?

Even when your URL is displayed as a source, the body of the answer is not necessarily correct.

• Old pricing is shown.
• A discontinued service is treated as current.
• Only one of several business lines is recognised.
• The regions served are described incorrectly.
• The representative or company information is out of date.
• The scope of the free diagnosis and of the paid work is mixed together.
• Another company’s track record is described as yours.

In these cases, check not only the site itself but also external profiles, press releases, social accounts, recruitment media, and partner articles.

Order of correction

• Correct the source of truth on the official site.
• Consolidate duplicate pages.
• Render fees, names, and target customers from shared data.
• Update external profiles.
• Re-observe with the same set of questions after an interval.
• Record the content and frequency of incorrect answers.

The answer does not necessarily change immediately after a correction. Retrieval and re-evaluation take time.

Why AI gets your company information wrong

Why your technical capability does not come across on the web

Stage 5: Does it lead to action on the site?

Even when an AI answer makes someone aware of your company, they are unlikely to proceed to an enquiry if the site gives them nothing to judge by.

• Whether the first screen says what the company helps with.
• Whether the target customer, and who is out of scope, are clear.
• Whether fees, or the conditions that determine cost, are clear.
• Whether the part you were responsible for in each case study is explicit.
• Whether the process after an initial conversation is clear.
• Whether the enquiry form is too long.
• Whether the boundary between free and paid is clear.
• Whether the contact, expected reply time, and required information are clear.

Do not let AI-search work end at off-site exposure.

How to diagnose why a B2B site is not generating enquiries

Diagnosis by symptom

When only competitors are recommended, do not simply copy their wording. State the conditions and deliverables that are specific to you.

To record the state of each stage, keep the generative AI service used, the question text, the date checked, whether your company was included, which page was cited, whether the company information was accurate, which other companies were shown alongside you, and the next page to fix.

How to measure AI visibility

Measuring Google AI features, and separating that from external AI citation tracking

Converting the diagnosis into an improvement specification

Convert the diagnosis into an improvement specification that has an owner and a deadline. The specification needs the following items.

• Target URL
• The fact observed
• Reason for the change
• Revised text
• Code changes
• CMS changes
• Release conditions
• How it will be verified
• The observation period after release

Until the improvement specification is implemented, what is published stays the same.

Seven reasons an SEO report never becomes an implemented improvement

How the five stages map to the SIGNAL diagnostic stages

The five stages in this article are a separation limited to visibility in AI search. The Netsujo SIGNAL diagnosis divides the whole sales foundation into eight first principles: existence, discovery, retrieval, trust, understanding, action, measurement, and improvement. They correspond as follows.

Measurement and improvement sit outside the five stages in this article. Because the free diagnosis works only from public information, analysis of the access data itself and evaluation of the improvement cycle are outside its scope.

What Netsujo SIGNAL checks

In the free Netsujo SIGNAL diagnosis, you first register four items: the service you most want to sell, the customers you want to win, the problem those customers have, and the action you want to increase. With those goals confirmed, we check the state of each stage from public information.

If the aim is to increase enquiries other than branded search, we do not evaluate using questions that contain your company name. We design non-branded questions along the stages of consideration: the problem, the way to solve it, the service options, the providers, the comparison criteria, and commissioning or adoption. Asking with your own company name will of course surface your company, and building the assessment on that makes non-branded discovery impossible to measure.

Stages that could not be observed are reported as not observed, not as no problem found. Where improvement is needed, we convert the observed facts into a specification that can be handed to a production company or an in-house developer.

The free diagnosis requires no card registration and does not roll over into a paid contract. It does not require a GA4 or Search Console connection either.

What an AIO diagnosis can and cannot tell you

Summary

A different blocked stage calls for a different response. Before adding more articles, confirm which stage has stopped.

See the Netsujo SIGNAL service

What AIO, GEO, and LLMO mean