How to tell whether ChatGPT cites your brand

This guide describes a measurement method. Automated AI prompt, mention and citation monitoring is not available in the current public BeKnow product. Collect the observations with an external service or a documented manual sample.

To learn whether ChatGPT cites your brand, one question such as “Do you know this company?” is not enough. You need questions a buyer might genuinely ask, a recorded configuration and a distinction between three outcomes: a name mention, a link to the site and a citation used as a source.

ChatGPT can search the web and provide source links, but not every answer uses search and not every mention has a citation. OpenAI also warns that search results and citations may be incomplete, outdated or incorrect. Monitoring should preserve observed answers and allow a person to verify the underlying pages.

Mention, recommendation and citation are different

A mention means the brand name appears. A recommendation means it is proposed for a need or compared with alternatives. A citation connects a statement to an inspectable source. A domain can even appear as a link without the brand being the primary recommendation.

The distinction changes the diagnosis. A brand may be known but not recommended, recommended without a link, or absent while one of its pages supports an answer. The guide to AI mentions versus citations explains the metrics.

Build a sample from real user questions

Do not use only prompts containing the company name. They measure assisted recognition, not discovery. Use a small set across the decision journey:

• educational questions about the problem;
• category questions asking which tools solve it;
• comparisons involving genuine constraints;
• practical questions such as why a competitor is cited;
• local or language-specific questions only for markets the business serves.

The guide to choosing AI visibility prompts shows how to avoid an unbalanced sample.

Run a sound manual check

Start a new conversation, use ChatGPT Search when you intend to observe web-grounded answers and record the exact prompt, date, language and visible mode. Do not add hints such as “include my brand”; that changes the behaviour being measured.

Save the full answer, not just a screenshot of the name. Open each citation and check that the page supports the statement. OpenAI documents that ChatGPT may search automatically or through the Search tool and that Sources can show citations and other relevant links.

One run is an observation, not market share. Responses may vary by session, model, location and time.

These workflows use the user's OpenRouter key. Model, frequency and prompt count affect provider usage. Start with a small sample, and never treat a run that did not execute as a zero result.

Inspect the cited pages

For every citation, record:

• domain and exact URL;
• source type, such as documentation, guide, comparison or forum;
• visible publication or update date;
• the statement it supports;
• competitors appearing in the same passage.

When page contents need systematic comparison, Firecrawl can retrieve observed sources. A repeatedly cited page is a reason for investigation, not proof of a universal ranking formula.

Make the site eligible without promising placement

OpenAI says that a site should allow OAI-SearchBot and its published IP addresses to be eligible for ChatGPT Search inclusion. Eligibility does not guarantee inclusion or placement; multiple factors influence results.

Important pages should also be accessible, canonical, current and able to substantiate their claims. Generic copy or repeated brand names do not automatically create authority.

Read the results consistently

Compare the same question over time and separate:

• brand present or absent;
• prominent recommendation or passing mention;
• domain link;
• owned page cited;
• competitors mentioned and cited;
• failed request or answer without search.

Do not count failures as absence. Do not silently change prompts or models. When the sample changes, create a new documented version.

Frequently asked questions

Does a mention prove ChatGPT read my site?

No. Without an inspectable source, you cannot attribute a mention to a particular page.

Can I guarantee a citation?

No. You can improve access and usefulness, but OpenAI does not guarantee inclusion or placement.

How many questions should I monitor?

Begin with a small set representing actual decisions. Expand only when you know how the extra observation will be used.

Should I track competitors too?

Yes. Brand absence becomes more informative when you know which alternatives and sources appeared instead.

Measure observed answers, not impressions

The useful question is not “Does ChatGPT know my brand?” but “For which questions does it mention, link or cite the brand, under which configuration, and who appears instead?” A stable sample turns isolated checks into an operational signal.

What BeKnow keeps

Keep interventions, hypotheses, assets, observations and decisions in a searchable workspace. Record what is known and what still needs checking.

A change in performance after an intervention does not prove that the intervention caused it. Cross-platform attribution and a complete analytics dashboard are not available today.

Project memory and data imports do not require an AI model key. A compatible external AI client may have its own costs. BYOK applies only to available functions that actually call an external provider.

Next step

Start with one project, one documented change and the evidence needed to review it. Source connections.

Record your first intervention. How it works.