The Citation Intelligence report is one of the most actionable views in LLM Metrix — and one of the most frequently misread. It shows which sources AI engines cited when answering your tracked prompts, and whether your own domain was among them. This guide walks through the panels the page actually renders and explains what to do with what you find.
What the Citation Intelligence report shows
At its core, the report answers one question: when AI engines retrieve content to answer queries in your category, which sources are they pulling from?
For your most recent scan, you can see:
- Every citation, one row per source, labelled with the engine that produced it and the prompt it answered
- Those same citations grouped by domain, with a count and the set of engines that cited each
- Whether your own domain was cited at all, and how many times
- The external domains cited instead of, or alongside, yours — optionally decorated with a domain-authority figure
Plus one series across scans: total citations, distinct domains cited, and citations to your own domain.
One classification, not four. The product splits cited domains into your own domain versus everything else. It does not label a cited domain as a competitor’s, a partner’s, a community’s or an aggregator’s — that classification does not exist anywhere in the data model, and a competitor you have configured by name is never joined to a domain in the citation list. If rival.com is cited, it sits in the same external bucket as nytimes.com. Making that distinction is your read, not the product’s.
This is different from your visibility score or mention rate — those measure whether your brand is mentioned. Citation intelligence measures whether your content is being retrieved as a source, which is the mechanism that produces mentions. For the underlying retrieval mechanics, see how AI engines cite sources.
What a citation is, and why it varies by engine
Worth settling before you read any of the panels, because it changes what the data can support.
Where an engine runs a genuine web-search tool — as Google’s Grounding with Google Search does, returning inline supporting links for the pages it retrieved — a citation is a real URL, and the row is a clickable link to the page. Where it does not, the “source” is a domain inferred from the wording of the answer, and the row shows the bare domain with nothing to click.
The first kind supports page-level analysis; the second only supports domain-level analysis. Perplexity and Google AI Overviews return real cited URLs on every call; the chat engines only do so when search grounding is enabled on the deployment, which is off by default. So on a typical deployment most of your citation rows are inferred domains, and you should not read them as evidence that a specific page was retrieved.
The stat cards
Four figures across the top, all describing the latest scan and not a period:
- Total Citations — every citation row in that scan
- Unique Domains Cited — how many distinct domains those rows span
- Engines Citing — how many of your engines cited anything at all. Treat this as the sample-size card: if only two engines cited anything, everything below rests on a narrow base
- Top Domain — the most-cited domain and its count
Citation volume over time
The one panel on the page that spans scans rather than describing the latest one. It charts three series — total citations, distinct domains cited, and citations to your own domain — one point per scan, and appears once you have at least two scans to draw.
Read it for direction, not detail. Own-domain citations climbing while total citations hold steady is the shape you want: the same retrieval budget, more of it spent on you. Total citations collapsing across the board more often means an engine stopped returning sources than that the web changed.
It is a volume series, not a per-page history. It cannot tell you that a specific URL dropped out of the results — a page that stopped being cited is simply absent from the latest scan’s rows, and nothing records that it used to be there. If you need that comparison, take it yourself from successive scans via GET /api/v1/scans, which returns full per-answer citation rows for up to 100 past scans.
Source gaps
The panel to open first, and the one that carries the actual gap.
It leads with a single status line: your domain was cited N times, or it was not cited at all. Being mentioned but never cited is a specific and common condition — the engines know who you are and send readers somewhere else — which is why it leads rather than hiding in a list.
Below that sits the list of external domains cited for your queries, each with a count and, where we have one, a domain-authority figure. That list is the outreach brief: these are the pages standing between your content and the answer.
What to look for:
- Is your own domain cited at all? A zero here outranks everything else on the page.
- Which third-party sources (review sites, trade publications, community forums) are being cited in your place — and does your brand appear on them favourably, unfavourably, or not at all?
- Any unexpected domains as sources: an old blog, a Product Hunt listing, a stale job posting.
Note that the panel renders counts, not percentages. If you want a share figure, divide by Total Citations yourself — and remember it describes one scan.
Top cited domains
The same citations grouped by domain and ordered by count, each row carrying the engines that cited it as badges.
The count is occurrences within one scan — across your prompts and engines — not repeated runs of one query. Each prompt is asked exactly once per engine per scan, so a domain showing ×9 was cited nine times across nine different (prompt, engine) pairs, not nine times out of ten runs of the same question. That distinction matters: ×9 spread across nine prompts means broad topical authority, while ×9 concentrated on one prompt answered by seven engines means one page dominates one question. The engine badges are what let you tell those apart.
The citation list, and filtering it
Below the search box, every citation is listed as its own row: the page title as a link where the citation carries a real URL, the bare domain where it does not, the engine that produced it, when it was cited, and the query it answered.
The search box is a free-text filter over that list — it matches the domain, the query and the engine name. There is no drill-in by query cluster, no per-query grouping and no rank number; type an engine name to see only that engine’s citations, or a domain to see every prompt it turned up for. The filter applies to this list only, not to the Source gaps or Top cited domains panels above it.
On ordering. Rows appear in the order each engine returned its sources, so the first source listed for a given answer is the one that engine put first. The product does not surface that as a rank number, does not average it, and does not chart it — treat the ordering as a soft hint about one answer, not as a metric you can track.
Third-party sources: what to do about them
Third-party sources — review sites, comparison pages, industry publications, forums — are often the majority of citations in any query set. These are pages you don’t control, but they’re surfacing to AI engines and potentially shaping your brand mentions.
Favorable third-party citations: Third-party pages that mention your brand positively and appear as top sources are working in your favor. Identify these and maintain the relationship with the publication or community that produced them.
Unfavorable third-party citations: Third-party pages that mention your brand negatively, inaccurately, or not at all while featuring competitors are working against you. Your options:
- For inaccurate content: reach out to request corrections
- For missing mentions: outreach to request inclusion in relevant comparison content
- For negative reviews: address the underlying product or service issues
High-authority third-party sources you’re absent from: If G2, Trustpilot, or a major industry publication consistently appears as a top citation source and your brand isn’t featured or highly rated there, that’s a specific, addressable gap. The review sites and AI visibility guide covers how to approach these placements.
Turning citation intelligence into an action list
Paid plans refresh on a weekly cadence, so one scan is roughly one week’s reading. A productive routine after each refresh:
- Check the own-domain line first. Cited or not cited, and how many times. It is one line and it is the whole headline.
- Read the volume trend for direction. Own-domain citations rising against a flat total is progress; a total that collapses is usually an engine change rather than yours.
- Work the Source gaps list top-down. The highest-count external domains are where the retrieval budget is going. Sort your outreach by count first, authority second.
- Scan the engine badges on Top cited domains. A domain cited by every engine is a structural gap; one cited by a single engine may just be that engine’s habit.
- Feed the content queue. External pages that keep turning up for prompts you should own go on the content gap backlog.
Two things this routine deliberately does not include, because the product does not support them: chasing a citation rank, and diffing this week’s URLs against last week’s. Neither is surfaced. If you want the week-over-week diff, pull successive scans from GET /api/v1/scans and compare them yourself.
Frequently Asked Questions
What’s the difference between citation intelligence and my visibility score?
Your visibility score and mention rate measure whether your brand is mentioned in AI responses. Citation intelligence measures whether your content is being retrieved as a source — the underlying mechanism that produces those mentions. Citation data tells you which sources are doing the work, so it’s more directly actionable for content decisions. Note that the two are computed independently: citations do not feed the visibility score.
Does the report show a citation rank?
No. Rows appear in the order the engine returned its sources, but no rank number is stored, displayed or tracked, and there is no ranking to trend. If you see a claim that a “citation rank” correlates with where your brand lands in the answer, it does not describe this product — brand position within the answer prose is a separate measurement, reported on the Rankings page.
Can I see which citations belong to a competitor?
Not automatically. The report classifies cited domains as yours or external, and nothing joins the competitors you configure to the domains in the citation list. You can read a competitor’s domain out of the external list yourself, but the product will not label, colour-code or group it as one.
My page isn’t appearing in the citation report at all — what’s wrong?
First check what it would take to appear: the report lists sources cited for your tracked prompts in the latest scan, so a page only shows up if an engine reached for it while answering one of those prompts. A page targeting a topic you do not track cannot appear no matter how well it performs.
If it should be in scope, the usual causes are that it isn’t reachable by the relevant AI crawlers (verify crawlability and your sitemap), or that it’s being passed over in favour of stronger sources. Bear in mind too that whether a page can be named at all depends on which engine cited you, not on a single deployment setting: Perplexity and Google AI Overviews return real URLs on every call, while the chat engines return a bare domain unless grounding is enabled (see the two kinds of citation row above). So if your domain is cited by a chat engine with no page named, that is the measurement mode rather than a verdict on the page — but if Perplexity cited your domain and named a different page of yours than the one you expected, that is a real finding about which page it found more answerable.
How should I use the third-party source section?
Identify favorable third-party pages that cite you positively and protect those relationships; flag unfavorable or inaccurate ones for correction or outreach. Pay special attention to high-authority sources like G2 or major publications that appear as top citations but don’t feature you — those are specific, addressable gaps worth pursuing.