Skip to main content
Every AI answer is assembled from sources. Source Intelligence shows you that pool: which sites AI platforms actually cite when answering questions in your category, how much of it belongs to you, and which sources are carrying your competitors instead. It is the off-site half of AI visibility. Site Health tells you whether your own pages are ready to be cited. Source Intelligence tells you where the citations are actually going — and most of them, for most brands, are going somewhere you do not control.

​
What the page shows

Citation Snapshot — your citation share over the selected window, and how it is moving. Citation Mix — the pool broken down by kind of source: your own domain, competitors, earned media, aggregators, community, and everything else. With a Per Platform Read underneath, because the mix is rarely the same on two platforms. This is usually the fastest read on the page: if your category’s answers lean on aggregators and you have no presence on them, that is the finding. Top Cited Sources — the hero table. Top Cited Pages — the same idea one level finer, at individual URLs rather than domains, filterable by format and category. Compare with competitors — cards for each tracked competitor. Opening one drills into how that competitor earns its citations. There is a card for you too, drilling into your own position the same way.

​
Reading the drill

This is the part worth slowing down on. Almost every “these numbers don’t add up” question resolves to one of the two ideas below.

​
Drilling shrinks the set you are measuring inside

A drill does exactly one thing: it narrows the set of answers under measurement. Every percentage at a level is measured against that level’s set — never the one above it.
So a percentage deep in a drill is not a slice of the number at the top. It is a fresh measurement inside a smaller set. The path you took is what defines the set, which is why the drawer states the scope you are standing in.

​
Some tabs are comparisons; others are compositions

The tabs do not all answer the same shape of question, and mixing them up is the single most common misreading. On Platforms, Topics and Prompts you are looking at a comparison. Both brands are measured against the same denominator, so “you 55%, them 81%” is a real 26-point gap, and the rows roll back up to the figure at the top of the drawer. On Sources and Pages you are looking at two compositions side by side. Your 12% and their 30% divide by different totals — your citations and theirs. A higher percentage in their column does not mean they are cited more on that page; it means that page is a larger fraction of their (possibly much smaller) total. Read each column downward to see what makes up that brand’s position.
Because Sources and Pages are compositions, do not read the two percentages against each other. The honest finding in a pair like “them 30%, you 13%” is usually about concentration: a third of their entire position on that source rests on a single page, while yours is spread across many. That is a real fragility signal about them — it is not a statement that they beat you there.

​
Two habits that prevent most misreadings

Read the volume column before the percentage. “208 answers on Google AI Overviews cite this source, 17 on ChatGPT” is usually the finding. The percentage refines it; it rarely is it. Compare a rate to the brand’s own average, not to zero. A competitor named in 31% of answers that cite a given source looks weak until you know they are named in 33% of answers overall — at which point that source is simply big, not favourable to them. The interesting cells are the ones that diverge from a brand’s own baseline in either direction.

​
A source someone owns reads differently

Drilling into a competitor’s own domain, their “named in” column will sit near 100% — of course an answer citing their blog names them. That column is close to tautological and carries little information. The column that matters there is yours: when their content gets pulled into an answer, are you still in it? A source you cannot win is still a source you can measure yourself against, and the posture is defensive rather than contestable. On a source neither of you owns — a community site, a review publication — both columns are live, the gap between them means something, and an absence is addressable.

​
How to act on it

  1. Start with Citation Mix. If your own domain is a thin slice, the work is off-site, and no amount of on-page optimisation changes the pool.
  2. Find the big sources you are absent from. Top Cited Sources ranked by volume, filtered to where your share is zero, is your target list.
  3. Check the per-platform read before committing. A source that dominates on one platform and is absent on another is a narrower bet than it first appears.
  4. Use the competitor drills to find the specific page, not just the domain. “We should get covered by this publication” is a wish; “this one comparison article supplies a third of their citations on this topic” is a brief.