Citation gaps explained

Last updated:

A citation gap is a web page that an AI assistant cites when it recommends one of your competitors, but has never cited when talking about you. It is not a score and not a grade — it is a list of specific places where a competitor is part of the conversation and you are not yet.

First, what counts as a citation

Every time an AI engine uses a page to write one answer, that is one citation for that page. Ask the same question tomorrow and it uses the page again, that is two.

Think of a student writing essays. Each essay lists the books it used. If one book shows up in the bibliography of thirty essays, that is clearly the book the student reaches for. Citations work the same way, and the count matters as much as the list: a page cited every single day is a page that is quietly deciding your category.

Where the numbers come from

They are not estimates. When an assistant finishes an answer, its own system reports the exact pages it used — the same links a shopper sees under the answer. That list gets recorded. Nobody asks the model to remember what it read.

Gap, win, and the line between them

The gap list is a to-do list, not a report card. Every row is somewhere you could go and earn a mention.

The same website can be on both sides

This surprises people. Reddit has thousands of separate threads. If AI cites seven threads while recommending you and five other threads only for competitors, Reddit appears as both a win and a gap — because what matters is each individual page, not the website as a whole.

The one thing a gap does not mean

A gap means the assistant has never cited that page for you. It does notmean your brand is missing from the page. We record which pages the engines cite and who they cite them for; we do not fetch and read the page’s own text. Your brand could already be mentioned halfway down an article that AI has simply never used in your favour.

That distinction matters when you act on the list — you might open a gap page and find you are already on it. That is still useful information: it means being listed there was not, by itself, enough.

A playbook for working through the list

How long before it shows up?

Slower than you would like. Getting added to a page is step one; the assistant then has to re-read that page and start using it. Expect weeks, not days, and judge it on the trend rather than any single day’s answer. The underlying mechanism — how AI decides which brands to recommend — is worth understanding before you invest in the outreach.

See what AI says about your brand

SeenScan puts your customers’ questions to ChatGPT and Gemini every day and shows you where you’re named, where a competitor is named instead, and which pages to go earn. Your first result is free and needs no card; daily tracking runs on a 7-day trial that does require one.

Start tracking free