AI search
ChatGPT optimization: get your business recommended in ChatGPT
When someone asks ChatGPT for a clinic, a supplier or a property developer, it often searches the web, reads a few pages and names a few businesses. We work on the pages it reads and the sources it trusts.
Where ChatGPT gets what it says about you
ChatGPT answers from two places. The first is what the model learned during training: a large snapshot of public text, months old, that changes only when OpenAI releases a new model. The second is live search. For questions about prices, opening hours, local businesses or “the best X in Y”, ChatGPT usually searches the web, reads a handful of pages and writes its answer from them.
The two behave very differently. Training data is slow and you cannot edit it. Search results are current, and they are where most of the work pays off: a page fixed this month can be read by ChatGPT this month.
The answer also changes from one run to the next. Ask the same question twice and you may get two different shortlists. That matters for how results are measured, below.
OpenAI’s three crawlers, and what blocking each one does
OpenAI documents its crawlers in its bot documentation and treats each robots.txt rule separately. Getting them wrong is the most common reason a good site never appears.
| User agent | What it does | If you block it |
|---|---|---|
| OAI-SearchBot | Crawls pages so they can be shown in ChatGPT’s search answers | Your pages are not shown in ChatGPT search answers, though they can still appear as plain navigational links |
| GPTBot | Collects pages that may be used to train OpenAI’s models | You signal that your content should not be used for training. OpenAI says this setting is independent of search |
| ChatGPT-User | Visits a page when a person’s request in ChatGPT needs it | OpenAI says robots.txt rules may not apply, since a person started the visit. It plays no part in search inclusion |
How ChatGPT search shows its sources
When ChatGPT searches, it places links inside the answer and offers a list of the sources it used. The reader sees your name and a page title, then clicks through to one specific page, often a service page or an article, not your home page. That page has to make sense to a stranger arriving mid-thought, and it has to offer a next step. We usually put a booking form there.
Many people never click. They read the recommendation, then search your name on Google or type your address. Those visits show up as direct or brand traffic. Links from ChatGPT search often carry a utm_source=chatgpt.com tag, which helps, but the full picture needs both. Our guide to tracking AI search traffic explains how we join them up.
What we work on for ChatGPT
Crawler rules and firewalls
We set robots.txt for OAI-SearchBot and GPTBot with you, then check that your host, CDN or security plugin is not quietly blocking them.
Pages ChatGPT can read
Key facts in plain text in the HTML, not inside images, sliders or content that only appears after scripts run.
A prompt map
The questions your buyers type into ChatGPT, written the way they talk: “I need a dentist in my city who speaks French and does implants, who do you suggest?”
The sources it cites
For each prompt we record which websites ChatGPT links to. Those directories, review sites and articles become our list for outreach and corrections.
Comparison-ready pages
People ask ChatGPT to compare. Pages that say who you serve, what it costs or how pricing works, and how you differ give it something true to repeat.
How the work runs, month by month
Fix access first
Crawler rules and firewall settings are checked in the first week. OpenAI says search systems adjust to a robots.txt change in about a day.
Rewrite the pages it lands on
The pages ChatGPT cites, or should cite, get answer-first copy, clear facts and a next step.
Work the outside sources
We correct listings and earn mentions on the sites ChatGPT already cites for your questions.
Follow visits to leads
ChatGPT visits are tagged in our tracking and followed to the form, the call and the booked appointment.
Frequently asked questions
Should I block GPTBot?
It is a business decision. Blocking GPTBot tells OpenAI not to use your pages for training, and OpenAI says it is independent of search. A publisher whose content is the product may block it. A clinic or a property developer usually wants the model to know who they are, so we tend to leave it open.
How fast does ChatGPT notice changes to my site?
Search is the fast part: once OAI-SearchBot has re-read a page, ChatGPT search can use the new version. What the model learned in training only changes with a new model, which is out of anyone’s hands.
Why does ChatGPT say wrong things about my business?
Usually because an old listing, a directory or an outdated page says them, or because there is too little about you and the model fills the gap. Correcting that is the subject of our AI brand perception service.
Do I need an llms.txt file for ChatGPT?
OpenAI’s crawler documentation does not mention it. It is cheap to add and some tools read it, so we publish one when it helps, but it is not what gets you cited. Our llms.txt guide covers the details.
How is this different from your GEO service?
GEO covers every assistant. This service is the ChatGPT part in more depth: OpenAI’s crawlers, how ChatGPT searches and cites, and the prompts people type into it. Most clients take the two together.
Does it work in French and Arabic?
Yes. ChatGPT answers in the language of the question, and it often cites different sources in each language. We build and test a prompt set for every language you sell in.
Last updated · By the Napexy team
Keep reading
Tell us where you want to grow.
A 30-minute call with a strategist. You get a plan in plain words and a fixed quote, whether or not you work with us.