AI visibility
Brand citations on ChatGPT vs mentions in other AI engines
A citation points to a page or source; a mention names the brand without one. Here is how to tell them apart engine by engine, with a filled-in example for ChatGPT and Perplexity.
If you are tracking a site's AI visibility and your name turns up in an answer, the first thing to settle is how a brand citation on ChatGPT differs from a mention in Perplexity, Gemini, Grok or Google AI. The rule is simple: a citation points back to a page or source, while a mention names the brand with no source attached. Read it engine by engine, because each answer is its own event. Ask ChatGPT to compare tools and have the reply cite one of your pages with a link, and you have a citation. Find your name in a Perplexity sentence with no link back to the source, and you have a mention. Those are two separate rows in your log, not two points in the same count.
Why a citation and a mention are not the same event
RankAI starts from a working distinction. An answer can contain a citation, when it points to a page or a source; a mention, when it names the brand without a source; or an absence, when the brand does not appear at all. Apply these three labels to each engine separately. RankAI's public pages name ChatGPT, Perplexity, Gemini, Grok and Google AI. Claude is not one of those five and is measured only after you connect it, and the same goes for Grok. Gemini and Google AI Overviews are also not the same observation: keep them on separate rows even when the engine names seem to overlap.
A filled-in example
To make the rule checkable, here are two log rows as we would set them up:
- ChatGPT — citation: the answer points to a page on the site as its source; in the log we write "citation" and name the page cited.
- Perplexity — mention: the answer names the brand but does not point to a page; in the log we write "mention" and attribute no source.
The two rows stay apart. They are not added into a single indicator and not compared as if they were the same phenomenon. The same holds for Gemini, Grok and Google AI, each with its own row and its own label.
How RankAI separates citations from mentions
What follows is a working proposal from RankAI, not a universal standard. It is meant to answer "how do you tell them apart" with steps you can check.
- Decide first whether the occurrence is a citation, a mention or an absence. Check: does the answer point to a page or a source? If yes, it is a citation. If it names the brand with no pointer, it is a mention. If the brand does not appear, it is an absence. Assign the label before you note the engine.
- Attach the label to the right engine, without merging. Check: every log row carries a single engine among ChatGPT, Perplexity, Gemini, Grok and Google AI. Gemini and Google AI Overviews stay separate observations; Claude and Grok are recorded only after you connect them.
- Keep the evidence for the citation. Check: for every citation we name the page or source it points to; for every mention we write "no source" explicitly. If you cannot name the source, the row is not a citation.
- Read the log engine by engine, not in aggregate. Check: the citation count for ChatGPT does not include mentions found on Perplexity, Gemini, Grok or Google AI. If a total mixes the two labels, the log needs fixing.
The point RankAI wants to make explicit is that the word "visibility" tends to flatten different events. A citation on ChatGPT is an event with a source: it says a page on the site was called up in that answer. A mention on Perplexity, Gemini, Grok or Google AI is an event without a source: it says the name came up, but not that a page was used as a reference. If you only look at whether "the brand appears," the two look the same; if you look at the source, they no longer do.
That is why we suggest a two-column log: label (citation, mention, absence) and engine. The label column answers "with or without a source?", the engine column answers "where?". Only by crossing the two can you say, for instance, that ChatGPT gave you a citation and Perplexity a mention, without turning the second into a missed citation. It is a reading choice, not a universal measurement: others might organize the log differently, but the source test remains the most direct way to keep the two events apart.
A checkable limit
The method applies only to the engines RankAI names publicly — ChatGPT, Perplexity, Gemini, Grok and Google AI — and only to the observations available. Claude and Grok are measured after you connect them, and Gemini and Google AI Overviews remain separate observations. An API response is not the same as a person's screen: if the log rests on an API response, say so, because it is not what a user sees. If a row carries neither the source nor the label "mention," it cannot be checked and should be left out of the count.
For the full picture on AI visibility and how to read observations engine by engine, see the in-depth page.