The short answer
ChatGPT answers from two places. Its model learned from training data up to a cutoff date. When it searches the web, it uses OpenAI's own index, built by OAI-SearchBot, plus third-party search providers. To appear in ChatGPT search, allow OAI-SearchBot in robots.txt, publish clear pages, and earn mentions on sources it trusts.
- ChatGPT mixes what its model learned in training with live web search results.
- OAI-SearchBot builds the search index, GPTBot collects training data, and ChatGPT-User fetches pages on request.
- Blocking GPTBot does not remove you from ChatGPT search.
- Answers vary between runs, so check the same questions more than once.
How does ChatGPT find information?
ChatGPT has two main sources. The first is its training data, a large body of text collected up to a cutoff date. Answers from training alone can be out of date.
The second is live search. ChatGPT can search the web when a question needs current facts, or when the user asks it to. It then reads results and writes an answer with sources.
OpenAI says ChatGPT search uses its own index plus third-party search providers. OpenAI runs separate crawlers for separate jobs. Check its bot documentation for current details.
| User agent | What it does | If you block it |
|---|---|---|
| OAI-SearchBot | Crawls pages for ChatGPT search results | Your pages may not appear as ChatGPT search sources |
| GPTBot | Collects content that may train OpenAI models | You opt out of future training; search is unaffected |
| ChatGPT-User | Visits a page when a user's request needs it | OpenAI notes robots.txt may not apply to these user-initiated visits |
What does ChatGPT tend to show?
When ChatGPT searches, answers usually include inline citations and a list of sources. Clicking a citation opens the page.
Without search, it may name brands from memory with no links. Those descriptions can be dated or simply wrong.
For buyer questions, answers often follow a pattern. Layouts change often, so treat this as a rough guide.
- A short list of options, with a sentence on each
- Criteria to weigh, such as price, location or reviews
- Links to the pages, reviews or lists it relied on
- Product cards with images for some shopping questions
How to make your site readable for ChatGPT
Start with robots.txt. Allow OAI-SearchBot if you want to appear in ChatGPT search. Decide separately whether to allow GPTBot for training.
This example allows search and user visits but opts out of training. Remove the GPTBot block if you are happy to be in training data.
- Check your CDN or firewall does not block OpenAI's bots before they reach robots.txt.
- Make sure key text is in the HTML, not only loaded later by JavaScript.
- Put direct answers near the top of service, pricing and comparison pages.
- Keep facts consistent across your site, directories and review profiles.
# ChatGPT search index User-agent: OAI-SearchBot Allow: / # Pages fetched when a ChatGPT user asks User-agent: ChatGPT-User Allow: / # Model training (optional opt-out) User-agent: GPTBot Disallow: /
OpenAI publishes IP ranges for its bots, so your host can allow the real ones and block impostors.
How to check what ChatGPT says about your brand
You can run a useful spot check yourself in about half an hour.
- 1
Write your buyer questions
List five to ten questions a real buyer would ask. For example: which bookkeeper in Bristol suits a startup?
- 2
Use a clean session
Use a temporary chat where possible. Memory and past chats can shape the answers you see.
- 3
Ask with and without search
Ask once normally, then again with web search turned on. Compare what changes.
- 4
Record the results
Note who is named, in what order, which sources are cited, and how you are described.
- 5
Repeat next week
Ask the same questions again. Answers vary, so look for patterns rather than one-off results.
Does WebValet track ChatGPT?
Yes. ChatGPT is one of three platforms included in WebValet Pro, with Gemini and Perplexity. Pro asks 25 buyer prompts every week and records mentions, citations and the competitors named.
WebValet is in early access, and tracking features are rolling out to Pro. The free check reviews readiness today, including your robots.txt rules for OpenAI's bots. It does not ask ChatGPT questions yet.
Questions people ask
Does blocking GPTBot remove me from ChatGPT?
No. GPTBot collects training data. ChatGPT search relies on OAI-SearchBot, which has its own robots.txt rule.
How long until ChatGPT reflects changes to my site?
OpenAI does not publish a timeline. Search answers can change once OAI-SearchBot recrawls your pages. Answers from training data only change when models are updated.
Why does ChatGPT describe my business wrongly?
It may rely on old training data or on third-party pages with outdated facts. Fix the facts on your site and key listings first. Then check again with web search turned on.
Can I pay to appear in ChatGPT's answers?
You cannot buy a place in ChatGPT's regular answers. Visibility comes from clear pages, crawler access and mentions on trusted sources. Check OpenAI's site for any advertising options.