How to see if my website is cited in ChatGPT: a page-by-page check
Start from your own URLs, not from prompts. A page-level method for finding out which pages ChatGPT fetches, which it cites, and which it ignores, with a free first pass and a tool-based follow-up.
Quick verdict
Most guides tell you to type prompts into ChatGPT and look for your domain. That works, but it starts from the wrong end. Start from your pages instead. List the ten URLs you most want cited, then check each one for three things: can ChatGPT's bots reach it, does ChatGPT fetch it, and does it show up as a source. Promptwatch lets you do all three for specific URLs with its Page Tracker and crawler logs, which is why it is our pick once the free checks stop being enough.
Step 1: pick the pages that should be cited
Open your analytics or your sitemap and choose ten pages. Favor pages that answer a question directly: comparison pages, how-to guides, pricing explainers, and glossary entries. Skip your homepage. ChatGPT rarely cites a homepage for a category question.
Write the question each page answers in plain words. "How much does X cost for a small team?" is a prompt. "Our pricing page" is not.
Step 2: check that the door is open
A page cannot be cited if ChatGPT's bots are blocked. Open your robots.txt at yourdomain.com/robots.txt and look for rules naming these user agents:
- OAI-SearchBot powers search results in ChatGPT.
- GPTBot collects content that may be used for model training.
- ChatGPT-User fetches pages when a user's request triggers a visit.
A blanket Disallow: / under User-agent: * blocks all of them. So do some CDN or firewall bot rules that never appear in robots.txt. If you want to appear in search answers, OAI-SearchBot is the one you cannot block. There is no bot called ChatGPTBot, so a rule using that name does nothing.
Step 3: ask the question and read the sources
Open a logged-out or temporary ChatGPT session. Ask the question from step 1, worded as a buyer would ask it. If ChatGPT searches the web, it shows sources. Open the sources list and look for your URL, not just your domain, since the page cited tells you which format works.
Repeat the prompt three times. A page that appears once in three is borderline, and a page that appears in every run is secure. Log the result next to the page.
Step 4: test with a site search
A quick supplementary check is to ask ChatGPT a question where you know your page contains the answer, then add your brand. If the answer paraphrases your page without citing it, the content is being used without credit, which usually means a better-structured competitor page is winning the link.
Step 5: read your own server logs
If you can access server or CDN logs, filter for the three user agents above. Look for status codes. A 200 means the bot got the page. A 403, 429, or 5xx means it did not. Many sites discover here that a bot protection rule has been quietly rejecting OAI-SearchBot for months.
This step is the least glamorous and the most useful. A page with no fetches cannot be cited, and a page fetched often but never cited has a content problem, not a technical one.
What your results usually mean
| Pattern | Likely cause | First fix |
|---|---|---|
| Bot blocked in robots.txt | Overbroad rule | Allow OAI-SearchBot and ChatGPT-User |
| 403 or 429 in logs | Firewall or rate limit | Allowlist verified bot ranges |
| Fetched, not cited | Page does not answer the question directly | Restructure around the question |
| Cited sometimes | Competing sources close in quality | Add specifics, refresh the page |
| Cited every run | Page works | Protect it, keep it fresh |
Where tools save you time
Doing this by hand for ten pages is a long afternoon. Doing it weekly is a part-time job. Promptwatch maps to each step.
The Page Tracker watches specific URLs, yours or a competitor's, and lists the responses and prompts in which they appear. Agent Analytics reads IP-verified crawler logs, so spoofed user agents are filtered out, and shows the crawl-to-citation path with a citation rate per page plus error tracking with failure explanations. That replaces step 5 entirely. Citation analytics adds page and domain citations with URL positions.
Gating matters here. Crawler logs start on the Professional plan at $245 a month. Essential does not include them. The free Explore tier tracks 10 prompts on ChatGPT only, which is enough to check your top pages against the sources panel.
Promptwatch also has free tools worth trying first: the AI Visibility Report runs a no-login scan, the Agentic Readiness Check looks at how prepared your site is for AI agents, and the AI Bots directory lists 440+ crawlers.
Who should pick which
| You are... | Do this |
|---|---|
| A small site with ten important pages | The manual steps above, once a month |
| Unsure whether bots can reach you | Check robots.txt and server logs first, or run the Agentic Readiness Check |
| Tracking pages weekly | Promptwatch Page Tracker on Essential or above |
| Seeing fetches without citations | Professional, so you can read the crawl-to-citation path |
| Comparing against rivals' pages | Page Tracker on a competitor URL |
If you want to compare trackers that offer page-level views, see the ranking. For most teams, the useful order is: free manual check, then Explore, then a paid tier only once logs matter.
One last check
After you fix a block or restructure a page, wait a few weeks before judging. Bots need to revisit, and citations need to accumulate. Put a calendar reminder in, rerun steps 3 and 5, and keep the log. A page moving from fetched-but-uncited to cited is the cleanest proof that a change worked, and it is the kind of before-and-after you can show a stakeholder without a slide deck.