How to See If My Website Is Cited in ChatGPT: The Page-by-Page Method
A page-by-page method for finding out whether ChatGPT cites your site, why a fast check misleads, and how to monitor the answer over time.
"Is my website cited in ChatGPT?" sounds like a yes or no question. It is not. Your homepage might never appear while a single guide earns citations every week. Your brand might be named in an answer without a link. A competitor might be cited for a topic you thought you owned. A useful answer is a map of pages and prompts, not a single bit.
This post gives you a method that starts with a free manual check and ends with ongoing monitoring. If you read our earlier piece on checking whether ChatGPT cites your website, this one goes deeper on the page-by-page view and on what to do when the answer is no.
The quick check, and why it misleads
The fastest check is to open ChatGPT, ask a question your page answers, and look at the sources. It is a fine sanity test. It has three flaws.
First, one answer is one sample. Run the same question tomorrow and the sources can differ. Second, your wording is not the engine's wording. ChatGPT often breaks a question into several background searches, called query fan-outs, and cites from those. Third, you only test questions you already thought of.
Treat the quick check as a smoke test only. If you see your site, good. If you do not, that tells you very little.
Step 1: List the pages that should be cited
Write down 10 to 20 URLs: your best guides, your product pages, your comparison pages, your pricing page. For each, add the question you would want it to answer. This pairing of page and question is the backbone of the method.
Step 2: Check access before content
Before you judge the content, confirm that the right bots can reach the page. Three ChatGPT user agents matter: GPTBot, OAI-SearchBot, and ChatGPT-User. Check your robots.txt for rules that block them, and check any firewall or bot protection that might return errors to them. A page that is blocked or erroring cannot be cited, however good it is.
If you have server or CDN logs, search for those three names. If the requests are missing for a page, the problem is discovery or access. If the requests are there but the page is never cited, the problem is more likely the content or the competition.
Promptwatch reads this directly through Agent Analytics, with real-time IP-verified crawler logs that filter out spoofed user agents. It supports log collection through Cloudflare, AWS CloudFront, Fastly, Vercel, Netlify, Akamai, Google Cloud CDN, a custom HTTP endpoint, or the WordPress plugin. Crawler logs start on the Professional plan.
Step 3: Pin each URL
This is where a one-off check becomes monitoring. Page Tracker in Promptwatch lets you pin specific URLs, yours or a competitor's, and see the prompts and responses connected to them. Add your 10 to 20 pages. After a few days you can see which are cited, for which prompts, and in which responses.
The competitor option is the underrated half. Pin the page that keeps getting cited instead of yours. Read it. Compare its structure, its length, its specificity, and how directly it answers the question. Most citation gaps become obvious within ten minutes of that comparison.
Step 4: Separate three kinds of presence
Not every appearance is equal. Keep these apart in your notes.
Cited with a link. The page appears as a source. This is what most people mean by "cited."
Named without a link. The brand appears in the text, but no page of yours is listed. Promptwatch tracks these as offsite mentions when a third-party page names you without linking.
Absent. Neither the page nor the brand appears.
Each calls for a different fix. Absent usually means content or access. Named without a link usually means your brand is known but your page is not the best source. Cited with a link means you are in the game, and the question becomes how stable the position is.
Step 5: Look at the shape over time
Citation behavior has a life cycle. A new page may take days to show up, peak, then fade. Citation Trends in Promptwatch reports first cited, ramp-up days, peak rate, and decay half-life per page, and classifies each citation by content type and source type. If a page has a short half-life, schedule a refresh. If it never ramps up, go back to the access check.
Step 6: Decide what to fix
Make a three-column list: page, finding, action. Typical findings are "blocked bot," "no fetches in 30 days," "fetched but never cited," and "cited then dropped." Typical actions are "fix robots.txt," "add internal links," "rewrite the opening to answer the question directly," and "refresh the data."
The Action Board turns that list into a standing queue. Its Discovery agent runs weekly and proposes up to five actions backed by your own data. For pages where the fix is new content, the Content Refresh feature proposes diff-style updates to existing pages, and the Content Agent on Business can plan and publish against gaps.
Common reasons a site is not cited
- The bots cannot fetch the page.
- The page answers a different question than the one asked.
- A competitor's page answers it more directly.
- The page is thin, outdated, or buried behind weak internal links.
- Most of the citations for your topic come from Reddit, YouTube, or review sites rather than from brand pages.
That last one deserves a note. If third-party sources dominate your topic, publishing more on your own site will not be enough. Promptwatch has Reddit, YouTube, and LinkedIn citation reports so you can see which platforms the engines trust for your prompts.
The recommendation
Do the quick check today, then move the real work into a tool that can pin pages and watch them. Promptwatch is built around exactly that pairing, with Page Tracker for the page-by-page view and Agent Analytics for the access side. The free Explore tier is limited to ChatGPT, which is all this particular question needs. When you want to see the crawler side, move up to Professional.