ChatGPT can use your website through three separate paths: as training data collected by GPTBot, as a live source retrieved and cited by OAI-SearchBot when search mode runs, and as context a user pastes or points it at during a conversation. Each path is independently checkable, and each fails for a different reason, so "does ChatGPT use my site" is really three questions with three answers.
Path one: the training corpus
GPTBot is OpenAI's training crawler. If it fetches your pages and your robots.txt permits it, your content is eligible to become part of future model training. Eligible is the operative word: a crawl guarantees nothing about influence, because training distills patterns across the whole corpus rather than memorizing your pages. Your site contributes most when its facts about your brand are echoed by other sources, making the association stick. Check this path in your server logs: search for the GPTBot user agent and confirm it is fetching your important pages, then audit robots.txt for accidental blocks.
Path two: live search citations
When ChatGPT runs a search, OAI-SearchBot fetches candidate pages as raw HTML, and cited pages directly shape the answer. This is the path you can win in weeks rather than training cycles, and it has one hard gate: OAI-SearchBot does not execute JavaScript, so content that only exists after client-side rendering is invisible to it. Test the path end to end by asking ChatGPT (with search enabled) a question your page answers well and seeing whether you appear in the citations; test the gate by viewing your page source and confirming the substance is in the raw HTML. If it is not, start with making your site readable to AI crawlers.
Path three: user-supplied context
Users can paste your URL or ask ChatGPT to look at your site directly. It is the smallest path and you cannot optimize your way into it, but it raises the stakes on accuracy: when a prospect says "summarize this vendor," your pricing page, docs, and About page become the pitch. Stale or contradictory pages get summarized as confidently as good ones.
The 20-minute audit
- Grep your server logs for GPTBot and OAI-SearchBot user agents and note which pages each fetches.
- Read robots.txt and confirm you are not blocking the bot you meant to allow; the full crawler roster explains who is who.
- View source on your five most important pages and check the real content appears in raw HTML.
- Run ten search-mode prompts your buyers would ask and record whether your domain shows up in the citations.
Check all three paths in one scan
A free scan shows whether ChatGPT actually mentions your brand on buyer prompts and includes an AI-crawler readability check, so you know if the bots can even read your site.
Frequently asked questions
How do I know if GPTBot has crawled my site?
Check your server or CDN access logs for the GPTBot user agent string. The log entries show which URLs it fetched and when. If GPTBot never appears, verify your robots.txt is not blocking it and that your host or firewall is not filtering it out before it reaches your pages.
Should I block GPTBot in robots.txt?
It is a trade-off between control and visibility. Blocking GPTBot keeps your content out of future OpenAI training, which also forfeits base-model brand presence. It is a separate decision from search: OAI-SearchBot handles ChatGPT's live citations, so you can block training while staying citable, or vice versa.
ChatGPT cites my competitor's page for questions we answer better. Can I take that spot?
Often, yes. Search-mode citations are re-decided at retrieval time, so a crawlable, server-rendered page that answers the exact question more directly and freshly than the incumbent can displace it. Confirm OAI-SearchBot can read your page, sharpen the answer near the top, and re-test over the following weeks.
Related answers