Guide · AT-R13REV. 2026-A

Being found by ChatGPT: how to check your visibility yourself

Five test questions help you examine mentions of your business and what is said about it. That way you keep the answers, the actual sources and the conditions of the test on record in a way others can follow.

By Jordan WichertSenior backend developerLast updated:

“Which electrician in the Vogtland is good?” or “Who can help me with a roof renovation?” are possible questions to an AI assistant. Its answer may contain company names and source links. Whether your business appears in it is documented for this specific question. How this investigation relates to classic search is explained in the guide “SEO, AEO, GEO: What the difference is”.

To get started you need suitable questions and access to the search system you choose. Functions and usage limits can differ depending on the service and account. This guide helps you interpret the observations and avoid typical measurement errors.

Preparation: record the test context

Start with a new conversation and document the search mode used. Check whether memories or personal settings are taken into account. Earlier conversations about the business can influence the context; being signed in alone does not make a test unusable, however.

A private browser window or a temporary chat is no proof of a neutral result. OpenAI describes different personalisation options for temporary chats. If appropriate, choose the unpersonalised variant and note the setting. Language and location also belong in the record.

If you are examining web search, record whether a search was actually carried out and which sources the answer shows. Open the links and check whether the linked page supports the statement. The OpenAI help page on web search explains sources and the influence of context. An answer without web search is considered separately.

The five test questions

Replace the trade and place with your own details. Start a new conversation for every independent question, so that a company named earlier does not prime the next test. If you compare several systems, keep a separate record for each and note the functions actually available.

1. The provider question: “Which roofers are there in the Vogtland?” Note which providers the answer names. A missing name does not yet prove that the system has no information about that business.

2. The recommendation question: “Which roofer in the Vogtland do you recommend?” Check how the recommendation is justified and which sources are actually linked. The justification in the answer is not yet verified evidence in itself.

3. The problem question: “My roof is leaking, who can help me in Plauen?” This is how real customers ask: with the problem, not the name of the trade. A mention shows that your business appears in this answer. Check the details and sources to see whether the recommendation is appropriate.

4. The brand question: “What do you know about [your company name] in [your town]?” Now it gets personal. Check the answer for accuracy: are the services, address and opening hours correct? Incorrect details can mislead prospective customers. Record the specific discrepancy and its possible source.

5. The comparison question: “What sets [your company name] apart from other providers in the region?” Examine whether the differences named are backed by verifiable information. Keep this brand question separate from the general questions that contain no company name.

The record: what to note down

Note the exact question, date, system and search mode. Add the personalisation settings used and the location. Save the answer together with the sources actually linked. Record separately whether your company name is mentioned, your own website is linked and a statement is factually correct. Mentions of competitors do not prove lost orders.

QUESTIONNAMEDYOU?1 · Provider questionBusiness A, B—2 · Recommendation questionBusiness A, B—3 · Problem questionBusiness A, you

The graphic shows a simplified extract with invented example answers. For your own comparison, the test conditions and sources belong with it. Repeat suitable questions at a fixed interval and after important changes. Even under similar conditions, answers can turn out differently.

Two typical measurement errors

Error one: an undocumented context. A follow-up question after a conversation about your company is a different test from a fresh provider question. Search mode, personalisation and location can also differ. These differences must stay visible in any comparison.

Error two: the everyday-word effect. If your company name is also an ordinary word, such as a surname or a place name, the systems like to mix in hits from entirely different contexts when you ask the brand question. That distorts the result. In that case, always test with the company name plus the town, never with the name alone.

And a general point: AI answers vary. The same question may be answered slightly differently tomorrow, and other users may see other answers. A single test is an indication, not a verdict. That is why, in our reports, we test every question across several systems and record the date and method.

What to do if you do not appear

First check whether ChatGPT is allowed to read your site at all: the AI crawler check shows that. Then examine the specific observation. For which questions is a mention missing, and which sources are used? Together with the content of your website, this gives you starting points for further checks:

Check public content. Are your services and company details described clearly on pages that can be reached? Structured data can add to this visible information. ChatGPT can also name a business on the basis of third-party sources; inclusion in Google’s index is not a precondition for that and does not prove a ChatGPT mention either. Problems in Google search are dealt with separately in the guide “Why can’t Google find my website?”.

Say the same thing everywhere. Name, address, telephone number and services should be correct on the website and in the external sources actually used. Correct discrepancies with the relevant source.

Provide answers. Content that answers real customer questions directly helps prospective customers decide. Add clear details on who the service suits and what the next step looks like.

Read the sources actually used. Check the details in linked directories, reports or other sources. Additional entries should have a useful purpose of their own. A recommendation cannot be guaranteed by the number of entries.

How we examine technical access, content and documented answers together is explained in Being found by ChatGPT. There you will also find the process of a commissioned review and implementation.

Common questions

Does the self-test cost anything?

You can use the guide yourself. Which search functions are available without payment or sign-up depends on the service you choose and its current usage limits. Check the terms there before the test. You do not need to book any additional service for our guide.

How often should I test?

Repeat suitable questions at a fixed interval and after relevant changes. Record the date, system, search mode and personalisation each time. Individual answers can vary even with the same questions. Several comparable observations help to put things in context, but do not yet prove the cause of a change.

My competitor is named and I am not. Why?

A single answer first shows only which names came up for this question under the chosen conditions. Check the sources that are actually linked and repeat the question with the context documented. Only together with the existing content and technical findings can you examine which changes might make sense.

Is the self-test enough, or do I need a proper measurement?

The self-test provides first observations and pointers to incorrect information. It does not yet explain why a system names your business or leaves it out. A deeper investigation combines documented questions and sources with a review of your website. The scope depends on the specific question you want to clarify.

+49 3744 4399760Mon–Fri 8:00–17:00
CallCallback

PARTS LIST · SHEET 1/1

What this page is made of

Values measured at build time. With JavaScript, your browser measures them again live.

ITEMPARTTYPESIZE
01index.htmlDocument≈ 29 kB
02styles.cssStylesheet≈ 20 kB
03scripts.jsScript≈ 13 kB
04archivo-variable.woff2Font≈ 58 kB
05ibm-plex-mono.woff2Font≈ 30 kB
06favicon.svgImage< 1 kB

Advertising or analytics cookies: none. Without your consent, the page loads nothing from third-party servers apart from Cloudflare, our delivery service. Fonts, scripts and images are hosted by us. That is how we build your IT, too.

OUTSIDE OFFICE HOURS

Feel free to call anyway

Our regular office hours are Monday to Friday, 8:00–17:00.

In the evening and at weekends the number is forwarded to us, and often someone will pick up. We cannot promise that, though.

If nobody answers, leave a short message on the voicemail: your name, your number and what it is about. We will call back as soon as we can, on the next working day at the latest.

The call costs nothing. If someone has to start work straight away outside office hours, the surcharge in the price list applies.

Call now+49 3744 4399760

Prefer to write? Use the contact form or e-mail [email protected]. We will read it on the next working day.

DIAGNOSIS · AT-K2

What do you need help with?

↵ SELECTESC CLOSECALL: +49 3744 4399760

Custom settings

Which services may load?

Strictly necessary

The choice you make here and your settings for colour scheme and motion. These are kept as small entries on your device and are never sent to a server. They are needed for the site to work and therefore cannot be switched off.