Why doesn't ChatGPT know my company?What to check before making changes.

What does “ChatGPT doesn't know my company” actually mean?

Start with the exact question and answer. Asking for a description by name is different from asking for the best supplier in a category. An assistant can describe a business when named without selecting it for an open-ended shortlist.

If your business is recognized but absent from supplier suggestions, investigate why AI recommends competitors instead of your company. Check the buyer criteria and the evidence behind the shortlist separately from recognition by name.

If the response describes a namesake, use our guide to AI confusing your company with another company. Here, the focus is a missing or insufficient description of your own business.

A blank answer is not evidence that your company is untrustworthy, that nobody has written about it, or that an AI provider has excluded it. Those explanations require evidence beyond the omission itself.

Separate model knowledge from web search

An AI model may lack relevant information or have information that is out of date. OpenAI's developer guidance treats those as reasons to supply better context, and distinguishes that work from changing model behavior. This is a general limitation, not a diagnosis of your company. Source: OpenAI, Optimizing LLM Accuracy.

Web retrieval gives the system another way to obtain information. OpenAI documents a web search tool that can retrieve current internet information and return citations. That API documentation explains the mechanism; it does not establish which tools were used in your ChatGPT conversation. Source: OpenAI, Web search for the API.

For ChatGPT itself, inspect the conversation's search activity and sources. OpenAI's ChatGPT guidance says search results and citations appear when web search is used on the web; workspace settings can limit availability. Source: ChatGPT, Web search.

A response is not a training-data audit

Do not translate “I don't know that company” into “your website was never in the training data.” The response does not provide the evidence needed for that conclusion. Likewise, a correct answer after you paste an About page demonstrates a result with supplied context, not prior recognition.

A website update and a model update are different events

If a test genuinely has no retrieval and you do not provide the revised text, it has no live route to read that edit during the test. Evaluate whether a search-assisted answer finds the revised page separately. Avoid treating a useful answer after browsing as proof that an answer without browsing has changed.

Test recognition before giving ChatGPT your website

Use separate conversations for the following checks so you can keep the supplied information distinct. A fresh conversation alone does not establish an isolated experiment: note any personalization, instructions, files or connected sources that could affect the answer.

First, ask about the company by name

Use the exact public trading name. This starter prompt is an editorial template, not a tested guarantee of model behavior:

What can you tell me about [company name]?
If you cannot identify the organization reliably, say so.
Do not infer its activity from the words in its name.

Record whether search was available and whether the conversation shows it being used. Asking an assistant not to browse is an instruction, not technical proof that retrieval was disabled. If you cannot establish the condition, label it “retrieval status unverified.”

If the name is ambiguous, run a separate check adding only the country. Record that addition. Do not silently replace the first result with the more informative prompt.

Then, ask for a web-assisted identification

In a separate conversation with search available, ask for current public evidence:

Search the web for [company name] in [country].
Identify its official website and describe its main activity.
Cite the pages supporting your identification.
If several businesses match, distinguish them rather than guessing.

Open the cited pages. Check whether they identify the correct organization and actually support the description. If the search still does not resolve the business, try supplying the official URL in another labeled check. That tests identification with a known website, which is a narrower task.

Keep a small evidence log

• The date, product and model label, if displayed.
• The exact prompt and complete answer.
• Search activity, cited URLs and any supplied website or files.
• Whether the company was identified correctly, with the facts used to judge that.
• Any changed conditions in subsequent checks.

Save incomplete and unsuccessful results too. If you later share a before-and-after example, include what changed in the prompt or tools as well as what changed on the website.

If web search cannot find your business, check access

Ask your site maintainer to inspect the actual public pages and access settings. A browser loading the homepage for you does not, by itself, establish that a search crawler received the same content.

Check the right OpenAI crawler

OpenAI distinguishes OAI-SearchBot, used for ChatGPT search, from GPTBot, whose crawled content may be used for model training. Their controls are independent: allowing search does not require allowing training. OpenAI recommends permitting OAI-SearchBot in robots.txt and allowing its published IP ranges for search access. Sites opting out may still appear as navigational links. Source: OpenAI crawler documentation.

Review the intended policy before changing a block. Ask for the effective crawler rule and relevant firewall or server logs, rather than a blanket instruction to allow every bot. A verified access fix is worth recording; it is not evidence that your company has subsequently been mentioned.

Check Google indexing separately

For Google, examine the homepage and relevant company page in Search Console's URL Inspection tool. Google's technical requirements cover crawler access, a working page returning HTTP 200 and indexable content. Meeting those requirements does not guarantee indexing. Source: Google Search technical requirements.

Do not use an indexed Google page as proof that ChatGPT will mention the company. These are separate observations. Keep the Google indexing result and the ChatGPT response in separate fields in your evidence log.

Verify fixes before requesting another crawl

After correcting a genuine Google access or page problem, an authorized Search Console user can request indexing for a managed URL. Google says a crawl request does not guarantee inclusion, and repeated requests for the same URL do not accelerate crawling. A sitemap helps Google discover URLs. Source: Google, Ask Google to recrawl your URLs.

Make your company easier to identify on its own website

Once you have checked access, review whether a visitor can establish the company's identity from the page itself. Our recommendation is to answer these ordinary questions in clear, visible text:

• What is the business called, including any relevant trading-name relationship?
• What does it sell or do, and for whom?
• Where is it based, and which locations does it serve?
• Which domain and public profiles belong to this organization?
• Where can someone verify contact details or other factual claims?

Use facts you can substantiate. If an old brand name still appears on a profile you control, explain the relationship where appropriate. If a directory contains an error, submit a correction with evidence and record whether the publisher accepted it. Do not describe a requested correction as a completed one.

Use Organization markup for accurate identity information

Google says Organization structured data can help it understand administrative details and distinguish an organization in search results. That is a documented Google use case, not proof that markup makes ChatGPT recognize a business. Source: Google, Organization structured data.

Have your maintainer check that the markup describes your own organization and agrees with the page. Google's guidelines prohibit misleading or hidden marked-up content and do not guarantee a rich result even when the markup is valid. Source: Google structured data guidelines.

Keep Google AI guidance within its stated scope

For supporting links in Google's AI Overviews and AI Mode, Google requires a page to be indexed and eligible for a Search snippet. Its guidance says no special schema or additional AI text file is needed. These statements concern Google's features; they should not be presented as a universal rule for every assistant. Source: Google, AI features and your website.

A useful editorial priority is to repair a specific gap you found: a missing description, conflicting identity details or a page the intended crawler could not access. Publishing more pages without identifying the problem makes a later result harder to interpret.

Retest recognition, citations and recommendations separately

Choose a small set of relevant prompts and keep their wording when comparing results. Record the displayed model, search condition and date again. If a setting changed, label the comparison accordingly.

Use the outcome that matches the original problem:

• Recognition: did the answer correctly identify the company by name?
• Search retrieval: did it find a relevant page about the correct business?
• Citation: did it link to evidence supporting its description?
• Recommendation: did it include the company for the particular buyer request?

These are useful review categories, not interchangeable measures of visibility. A correct search-assisted description after a website edit is an observation to preserve. By itself, it does not show that the edit caused the improvement or that future customers will get the same answer.

Set your own review schedule around meaningful changes rather than repeatedly asking until you get a favorable response. Keep the unfavorable answers and state how many checks you actually ran if you report a result.

What ReputationScan can tell you about company recognition

ReputationScan currently tests Claude and Gemini models, not ChatGPT. Use the manual checks above for the ChatGPT response that prompted your question. A result from a different provider does not measure that response.

The scanner asks its tested models about your company by name, with the country when supplied. Those queries have no web tools and do not receive your website. The website and public identity signals are collected separately. This separation is explained in our scan methodology.

The recognition check counts which successful model responses report recognizing the company. It is based on model self-report, not an inspection of training data. Read the actual descriptions and the models that returned results alongside the label. See what the scan checks.

ReputationScan refers interested businesses to SecondSideMedia, a partner that publishes corrective public records, and may earn a commission from resulting business. A scan does not change public records or model answers.

Sources and review notes

Documentation checked on 13 September 2026. The prompts, evidence log and diagnostic sequence are editorial recommendations, not experimental results. Product descriptions were also checked against the scanner implementation.

1. OpenAI — Optimizing LLM AccuracyMissing or outdated context and the distinction between context and model optimization.
2. OpenAI — Web search for the APIRetrieving current web information and returning citations.
3. ChatGPT — Web searchSearch activity, citations and workspace availability.
4. OpenAI — Overview of OpenAI CrawlersSeparate search and training controls; access guidance.
5. Google — Search technical requirementsAccess, successful page responses and indexable content.
6. Google — Ask Google to recrawl your URLsIndexing requests, sitemaps and the limits of recrawl requests.
7. Google — Organization structured dataOrganization details and disambiguation in Google Search.
8. Google — General structured data guidelinesAccurate, visible content and rich-result eligibility.
9. Google — AI features and your websiteRequirements and guidance for Google's AI search features.