DATA SHEET DS-01-C · CHATGPT VISIBILITYREV. 2026-A
Have your website optimised for ChatGPT
Some customers now ask ChatGPT instead of Google: “Which electrician in Auerbach also installs wall boxes?” The answer names a few businesses. Yours is not among them, or it is and the opening hours are wrong. We set up your website so that ChatGPT is allowed to read it, understands what you offer and can quote from it.
That means three jobs. We open access for OpenAI’s search crawler. We describe your services so that an assistant can find a clear answer there. And we use real questions to check what ChatGPT says about you before and after. We start with our free AI visibility check. Phone: +49 3744 4399760, Mon–Fri 8:00–17:00.
01 / WHAT WE DO
ChatGPT names other businesses and not yours?
Then either ChatGPT cannot reach your website, or it finds no sentence there that it can use as an answer. Often it is both. We start in three places.
Access: the crawler has to get in.
For ChatGPT search, OpenAI sends out its own crawler, OAI-SearchBot. If your robots.txt turns it away, or the host’s firewall takes it for an attacker, your website stays invisible to search. We check whether it gets through and clear the way. We keep the training crawler out if you want that.
Content: the page has to contain an answer.
An assistant is not looking for attractive pages. It looks for sentences that answer a question: who does what, for whom, where and on what terms. We rewrite your service pages so that those sentences are there, and make sure your name, address and services are the same everywhere.
Answers: check what ChatGPT says now.
Before we start, we ask ChatGPT the questions your customers would ask and write down the answers. Afterwards we ask the same ones again. That way you can see the difference and do not have to take our word for it.
If you would rather try it yourself first: our guide to the ChatGPT self-test shows how. This page describes what we change on your website.
02 / WHAT IT IS ABOUT
Named, linked, visited: three different things.
“Appearing in ChatGPT” can mean three things. ChatGPT names your company in the answer. Or it links one of your pages as a source. Or someone clicks that link and lands on your site. One does not follow from another: your name can come up because a trade directory lists you, without your website ever having been read.
For the work, this makes a difference. If the mention is missing, it is about content and company details. If the link is missing although you are named, ChatGPT is evidently reading about you somewhere else rather than on your site; then we look at access. And if no visitors come despite a link, it comes down to what the answer says about you.
OpenAI describes in its web search help (in English) that ChatGPT searches the web for current questions and links its sources. These source links are our most important tool. They show which page is behind a statement.
03 / CRAWLER ACCESS
Can ChatGPT reach your pages at all?
OpenAI is out on the web with three different access routes, and each has a different job. OAI-SearchBot collects pages for ChatGPT search. GPTBot collects content that OpenAI may use to train its models. ChatGPT-User fetches a single page when a user asks for it in the conversation, for instance by pasting in your address.
For being found, the first one counts. According to OpenAI, search and training can be controlled separately: you can let OAI-SearchBot in and keep GPTBot out. OpenAI publishes the names, jobs and IP ranges of the three in its crawler documentation (in English). We use it to check whether a request in the server log really comes from OpenAI.
We read your robots.txt and check which rule applies to which of the three. Then we fetch your most important pages the way a crawler does. A page that looks perfect in the browser sometimes answers a crawler with a security check, a login page or nothing at all. If the block sits with the host or in an upstream protection service, we talk to them and get an exception added.
What is meant to be protected stays protected. We only open what your customers are allowed to see anyway.
04 / QUOTABLE CONTENT
Is there a sentence on your page that ChatGPT can take as an answer?
Think of the question a customer types: “Is this right for my business?” An assistant then looks for a page that answers exactly that. If all it finds on yours is “We offer tailored solutions for every need”, it has nothing to work with and takes someone else’s page.
So we rebuild your service pages to a fixed pattern. At the top is the answer in a few sentences: what you do, for whom, in which area. Below that the requirements (from what size, for which devices, with what lead time). Then the evidence: examples, proof or figures that you can really show. And at the end, a way to get in touch that works.
We are not writing for machines alone. A customer who lands on your page through a single question and does not know the rest of the website needs exactly the same thing.
Where a question deserves a page of its own, we create one. How this fits into the whole website is covered under web design and development.
05 / COMPANY DETAILS
What else the web says about your business.
The same name, the same address, the same number everywhere.
ChatGPT does not only read your website. It also reads trade directories, map profiles and old listings that nobody has touched for years. If a former phone number or a service you stopped offering long ago is listed there, it can end up in the answer. We collect the listings, compare them with your details and put right whatever you have access to.
Location is not the same as service area.
If you work from Auerbach but also in Plauen and Zwickau, you have no branch there. The website has to keep both clearly apart: where the business is based and where it travels to. Otherwise an assistant can easily conclude that there is a branch that does not exist. Our own location is Rodewisch; how to describe a service area honestly is covered in our guide Web design in the Vogtland.
06 / WRONG INFORMATION
ChatGPT says something wrong about your company?
Old opening hours. A service you discontinued years ago. A business with a similar name in another town whose details end up with you. It happens, and there is usually a source.
We open the links that ChatGPT gives under its answer and look for the place where the wrong information appears. Then it depends on who owns the page. If it is your own website, we correct it. If it is a profile that you manage (map profile, business directory), we correct it using your access. If it is someone else’s page, we write to the operator and point out the spot; they have to change it themselves.
After that we ask the same question again. A correction reaches the answers with a delay, so we ask once more later.
07 / SCHEMA, LLMS.TXT, MCP
Schema, llms.txt, MCP: which of these does your website need?
Structured data (Schema.org): yes.
These are invisible labels in the source code that tell a machine: this is the company name, this is the address, these are the opening hours. We add them on every page where they fit, and in such a way that they match the visible text word for word. Labels that say something different from the page do more harm than good.
llms.txt: optional.
A text file that lists the most important pages of a website and describes them briefly, intended as a table of contents for language models. OpenAI does not list it as a requirement for ChatGPT search. We create one if the website is well organised and someone keeps the file up to date; an out-of-date llms.txt is worse than none. Ours shows what such a file looks like.
MCP interface: a different matter.
Through MCP, an assistant accesses data or functions directly, for example available appointments or stock levels. That is a development project of its own with its own access rights. It has nothing to do with whether ChatGPT names your business in search. If you need something like that, see MCP integrations.
08 / RE-CHECKING ANSWERS
The same questions before and after. That is how you see what has changed.
Before we change anything, we agree the questions with you. A few that contain your company name (“What does company so-and-so do?”) and a few open ones, the kind a new customer would ask (“Who repairs bean-to-cup coffee machines in Plauen?”). Each question gets a fresh conversation, so that the previous answer does not colour the next.
For each answer we note: the date, whether web search was active, which location the question was asked from and whether personalisation was switched on. Then the main point. Are you named? With what source? Is the information correct?
After the work we ask the same questions again under the same conditions and lay the two logs side by side. The conditions are no minor detail. Even a “temporary chat” is not neutral by itself: according to the OpenAI help on temporary chats (in English), it can be personalised depending on the settings. That is why the setting goes into the log.
09 / FREE FIRST STEP
Two ways to see where you stand. Both are free.
The AI visibility check: what do the assistants answer?
You tell us your business and sector. We ask five AI search systems about you the way a customer would, and send you the log with our comments: where you are named, which sources are behind it, what is wrong.
From that you can see whether and where the work is worthwhile. We describe support across all systems, with regular re-measuring, under AI-SEO and GEO.
The SpotLens check: what does a machine see on your home page?
SpotLens is our own testing tool, which we develop together with Arrowtrade. The free quick check looks at your home page from the outside and shows whether anything technical is in the way. The full audit goes deeper and sorts the points by urgency; what it includes is described on our SpotLens website audit page.
10 / HOW IT WORKS
From the first log to the changed website.
Asking
We put your customers’ questions to ChatGPT and write down what it answers today. You tell us what is correct.
Opening access
robots.txt, firewall, host: we check whether the search crawler gets through and fix whatever holds it up.
Revising pages
The service pages get clear answers, the company details are brought into line, and the markup in the source code matches the text. A change only goes live once you have approved it.
Re-measuring
The same questions, the same conditions, a second log. You see side by side what has changed.
Staying on it, if you wish
Answers change. If you want to keep an eye on that, we re-measure at regular intervals.
11 / SCOPE AND QUOTE
We start with the website you have.
Often it comes down to a handful of pages and a few settings. We only suggest a rebuild if the existing website cannot reasonably be reworked, and then with reasons.
If someone else looks after your website, they get a list from us that they can act on. Or we talk to them directly.
And if it turns out on closer inspection that you are missing from Google too: that is a different job with different tools. We describe it under Website not found on Google.
Specification
DATA SHEET DS-01-C · CHATGPT VISIBILITY
- Access
- ROBOTS.TXT · OAI-SEARCHBOT · FIREWALL AND HOST
- Content
- SERVICE PAGES WITH A CLEAR ANSWER · COMPANY DETAILS WITHOUT CONTRADICTION
- Markup
- SCHEMA.ORG MATCHING THE VISIBLE TEXT · LLMS.TXT IF NEEDED
- Verification
- THE SAME QUESTIONS BEFORE AND AFTER · LOG WITH SOURCES
- Starting point
- YOUR EXISTING WEBSITE
- Getting started
- FREE AI VISIBILITY CHECK ON REQUEST
- Quote (Angebot)
- IN WRITING BEFORE WE START · INITIAL CONSULTATION FREE
Questions & answers
Common questions, answered briefly.
We have a decent website. Why does ChatGPT still not mention us?
There are three usual reasons. OpenAI’s search crawler cannot reach the site because robots.txt, a firewall or the host turns it away. Or it gets in and finds no place that says clearly what you do and for whom. Or other sources on the web (trade directories, old listings) say something different about you than your website does. We check which of these applies to you.
I gave ChatGPT my address and it could read the page. Is that enough?
Unfortunately not. When you give it the address, ChatGPT fetches that one page, and it does so through a different access route from the one used for search. Whether it finds your business on its own when someone asks “electrician Auerbach wall box” is a separate question. We test that separately.
Do I have to allow OpenAI to train on my texts?
No. OpenAI separates the two itself: OAI-SearchBot handles search, while GPTBot collects content for training its models. You can let one in and keep the other out. We set it up the way you want it.
Do I need an llms.txt?
Not necessarily. OpenAI does not list it as a requirement for ChatGPT search. It does no harm and is quick to create if the website is well organised. The work that matters is in access and content.
Does my website have to rank well on Google first?
No, they are two separate systems with their own crawlers. But a lot helps both: pages that can be reached, say clearly what is offered and carry the same company details everywhere. If you are missing from Google as well, we will look at that too.
ChatGPT gives the wrong opening hours for us. Can you change that?
Not in ChatGPT itself, but at the source. We check where the information comes from: your own website, a profile you manage yourself, or someone else’s directory. We correct the first two. For the third we write to the operator; they have to make the change themselves.
Does this also apply to Perplexity and other AI assistants?
The groundwork does: reachable pages, clear statements, consistent company details. But every provider has different crawlers and rules. This page describes the work for ChatGPT. Looking after several AI search systems at once is our AI-SEO and GEO service.
Do we need a new website for this, and what does it cost?
Very few businesses need a new website. We start with what is there and change individual pages and settings. The AI visibility check at the start is free, and so is the initial consultation for business clients. For anything further you get a written quote (Angebot) before we begin.
When will ChatGPT mention us, and can you promise that?
OpenAI decides which businesses ChatGPT names, and the answers change from one question to the next. So nobody can promise you a mention, and neither can we. What we do: open up access, sharpen the content, put your company details in order and re-check with the same questions to see what has changed. What is most likely to improve is whatever depended on your own website: the crawler gets through, and the information about your business is right because the source is right. The log shows whether you appear when people ask open questions about providers.
Contact
For which questions should ChatGPT name your business?
Tell us your web address and your sector. For the free AI visibility check we ask five AI search systems about your business and send you the annotated log. For anything further you get a written quote beforehand.