AI visibility
How to Tell If ChatGPT Cites an Inner Page, Not the Home Page
Checking whether ChatGPT cites your brand is not enough. You need to see which URLs come back. Here is how to read the distribution of cited pages, and where the data stops.
Say you search for your brand on ChatGPT and the answer cites a blog post rather than your homepage. To notice that, it is not enough to check that the brand name appears. You have to look at which URLs come back as sources. The useful answer sits in the distribution of cited URLs: if only the home page shows up, citations are concentrated there; if inner pages show up, visibility is spread across several addresses. RankAI makes that distribution readable for the public pages it names, so the difference between home page and inner page becomes something you can observe.
Read the distribution of cited URLs, not just whether the brand appears
The question is about brand citations on ChatGPT, but the point is narrower: not how often the brand appears, but which pages get cited. An answer that names the brand and links to the home page tells you one thing; one that links to a product page or an article tells you another. The distribution of cited URLs is therefore the right unit to read: group citations by address and see where they cluster.
A working method proposed by RankAI
What follows is a working proposal from RankAI, not a universal standard. It is meant to help you read the data without confusing the presence of the brand with the citation of a specific page.
- Group citations by URL. Testable rule: every citation must be tied to a precise address, not to a generic “brand site.” If the address is not recorded, the measurement should be marked N/A.
- Separate the home page from inner pages. Testable rule: the home page has a single address; anything that does not match that address is an inner page. Counting the two groups separately makes the shift visible.
- Compare the same question over time. Testable rule: the same query, repeated, should produce a comparable distribution. If a measurement is missing in one run, it should be shown as N/A, never as zero.
- Check which engine the citation came from. Testable rule: a citation seen on ChatGPT is not automatically a citation on Perplexity or Gemini. The public pages name ChatGPT, Perplexity, Gemini, Grok and Google AI; Claude and Grok are measured only after you connect them.
What RankAI shows, and what remains a limit
The public pages name ChatGPT, Perplexity, Gemini, Grok and Google AI. When a visibility figure is unavailable, the product shows it as N/A, never as zero. That keeps you from reading a missing data point as a missing citation. Two limits remain. First, what comes back from an engine through an API is the API response, not the screen a person sees; the URL distribution has to be read within that boundary. Second, Gemini and Google AI Overviews are not the same observation, so they should not be added together as if they were.
The practical contribution here is to move the question from “is the brand cited?” to “which address is cited?” The first is answered with a yes or a no; the second needs a table where each row is a URL and each column is a run. In that reading, the home page is just one possible row and gets no special treatment: if an inner page collects more citations, the distribution shows it without any interpretation. So the method RankAI proposes is to treat the home page as one URL among others, record the exact address of every citation, and let the count show where visibility clusters. When an address cannot be reconstructed, that row stays N/A: a declared gap is more useful than an invented zero, because it signals that part of the distribution was not observed.
One limit you can check: if the engine does not return the address of the cited page, the distribution cannot be attributed with certainty and RankAI shows it as N/A. In that case you cannot claim that an inner page was cited instead of the home page.
To place this reading in the broader picture of AI search visibility, see the in-depth page.