Skip to main content
The citations report shows which URLs are being used as sources in AI answers — yours and everyone else’s — for your tracked prompt set.

​
What the report contains

Your cited URLs. Every page of yours that appeared as a source, with how many times and for which prompts. This tells you which pages are actually working. Cited domains overall. Every domain appearing as a source across the prompt set, ranked by citation count. This is the competitive field as the engines see it, which often differs from the one you track commercially. Per-prompt breakdown. For each prompt, the sources the answer drew on. The most useful view for diagnosis. Engine split. The same prompt across ChatGPT, Perplexity, AI Overviews, and Claude, since results differ substantially by engine.

​
How to read it

Start with the prompts where you are absent. Look at what was cited instead. The gap between those pages and yours is the actionable finding — usually a page that answers the question more directly than anything you have published. Look at page shape, not just domain. When competitors are cited, note which page. It is usually a narrow, factual page — one feature explained, one setting defined, one problem solved — rather than a landing page or a long guide. Check whether your cited pages are the ones you intended. Frequently the cited page is not the commercial page you care about. That is still useful, but it means the answer is not routing buyers where you want them. Read the domain distribution for structure. If one platform’s own domains take most citations for a cluster of prompts, that cluster has a ceiling. Do not spend content effort there — see prompt monitoring.

​
Common patterns and what they mean

​
Interpreting counts honestly

Citation counts are small numbers and behave like small numbers. A move from 3 to 5 is not a 67% improvement; it is noise unless it holds across several cycles. The reliable read is direction across a quarter, on a stable prompt set, against a recorded baseline.