GEO
ChatGPT visibility: be named when buyers ask
ChatGPT visibility work starts with a question almost nobody checks: whether OpenAI can fetch your pages at all. ChatGPT reaches the web through a different crawler than the one that trains its models, the two are controlled separately, and blocking the wrong one removes you from answers while doing nothing to protect your content.
Who it is for: Businesses whose buyers open ChatGPT before Google, and who have never checked what it says about them.
Everything in this service
- A robots.txt audit that separates the training crawler from the retrieval one
- Verification that OAI-SearchBot and ChatGPT-User receive real HTML, not an empty shell
- A fixed prompt set built from how your buyers actually phrase the problem
- A logged baseline: named or not, described how, and which sources the answer drew on
- Page-level work so the claims about you are specific enough to be worth quoting
- Monthly re-runs of the same prompts, with the wording of each answer recorded
What to expect
- A clear answer on whether OpenAI can currently read your site at all
- A written record of how ChatGPT describes you, and whether that description is accurate
- Pages that give the model something specific to quote rather than something vague to skip
How chatgpt visibility actually works
The crawler split, and why it costs people citations
OpenAI documents more than one agent, and they do different jobs. GPTBot collects content that may be used to train models. OAI-SearchBot surfaces and links websites inside ChatGPT search. ChatGPT-User fetches a page because a person asked about it in a conversation.
They are named separately in robots.txt, which means the decision is not all or nothing. You can decline model training and remain eligible to be cited and linked. Plenty of businesses would choose exactly that if anyone told them it was an option.
What usually happens instead is a single rule aimed at "AI scraping" that catches every agent with AI in the name. The content stays out of training, which was the goal, and the business also disappears from the answers that would have sent it customers. Six months later the conclusion is that AI search does not work for them.
- GPTBot: training. Blocking it keeps your content out of model training
- OAI-SearchBot: retrieval. Blocking it removes you from answers that link out
- ChatGPT-User: fetches a page because a person asked about it directly
- Names change. Check the vendor documentation rather than a blog post
What makes a page worth quoting
A model assembling an answer is looking for a passage it can lift and stand behind. That is a higher bar than being readable. A sentence that only makes sense with the paragraph above it is risky to quote, so it tends to be skipped in favor of one that does not need the context.
The practical test takes a minute. Take any paragraph off the page, read it cold, and ask whether it is still true and still complete. If it is, it is quotable. If it needs a run-up, rewrite it so the claim comes first and the reasoning follows.
Specificity does the rest. Vague claims are not worth quoting because quoting them commits the model to nothing. A named process, a real constraint, an actual number or a stated limitation all give it something to attribute.
Measuring this without fooling yourself
One check is an anecdote. The answer varies between accounts, sessions and phrasings, so a single screenshot proves very little and is easy to cherry-pick. A fixed prompt set, re-run on a schedule and written down, is a measurement.
Record three things each time: whether you were named, the exact wording used to describe you, and which sources the answer drew on. The description matters as much as the mention, because being named inaccurately is a fixable problem and you will not fix it if nobody is reading the answers.
Expect the wording to shift before the citation does. A description that moves from vague to specific is progress even in a month where you were not named, and it is the earliest signal available that the input layer is improving.
A clear path, step by step
- 01
Check the gate
We confirm which OpenAI agents your server admits and what they receive. This is where most invisibility turns out to live, and it is usually a five-minute fix once found.
- 02
Ask the questions your buyers ask
We build a prompt set from real buying language, not from your brand name, and record what comes back before anything changes.
- 03
Make the claims quotable
We rewrite the pages that should be answering those prompts so each claim stands alone, carries a specific detail, and survives being lifted without its surrounding paragraph.
- 04
Re-run and compare
The same prompts, monthly, logged. Movement shows up as a change in wording long before it shows up as a citation.
Why choose us for this
We check the crawler split first, because a blocked retrieval agent makes everything else pointless
We test with buying language, not brand names, so the result reflects how you would actually be found
We will tell you when the honest answer is that your buyers are not asking ChatGPT
Common questions
I blocked AI bots to protect my content. Have I blocked myself out of ChatGPT?
Possibly, and this is the most common self-inflicted problem in this area. OpenAI runs separate agents for training and for answering. GPTBot is the training crawler. OAI-SearchBot is the one that surfaces and links sites inside ChatGPT search. A blanket block catches both, so the content stays out of training and the business also stops being linked. If the goal was to decline training, the fix is to block one and allow the other.
Source OpenAI: Bots and crawlers (opens in a new tab) “GPTBot is used to make our generative AI foundation models more useful and safe.”
Source OpenAI: Bots and crawlers (opens in a new tab) “OAI-SearchBot is used to surface websites in search results in ChatGPT's search features.”
Can you guarantee ChatGPT will recommend my business?
No. The model chooses, the answer changes between sessions and phrasings, and no agency can prove a specific mention was caused by its own work. What is controllable is whether you can be read, whether there is anything worth quoting on the page, and whether the description of you is accurate. We do those and measure the rest.
How do I check what ChatGPT says about my business right now?
Ask it the question a buyer would ask, not your company name. "Who should I hire to do X in Y" tells you something. "What is Acme Ltd" only tells you whether it has heard of you. Run several phrasings, because one result is an anecdote. Write down whether you were named, how you were described, and which sources the answer cited.
Is this the same as ranking in Google?
Related but not identical. The same foundations matter: reachable pages, real answers, a consistent entity. What differs is that ChatGPT reaches the web through its own agents on its own schedule, so a page can rank respectably in Google and still be invisible here if the retrieval crawler is blocked or the page needs JavaScript to show its content.
Explore related work
Want this for your business?
Book a free visibility call and I will tell you honestly whether I can help.
How this is delivered
One person leads every project. Where a job genuinely needs a specialist, I bring in people I have worked with before and manage them, so you get one point of contact and one invoice rather than three suppliers blaming each other.
- You talk to the person responsible for the work, not an account manager
- Specialists are briefed and managed by me, and their work is checked before it reaches you
- One contract, one invoice, one place to chase