Skip to content
RankSmith

Citation intelligence

Knowing your client was mentioneddoes not tell you what to change.

An answer is assembled from sources. Until you know which pages the model drew on, every fix is a guess. The most common one, publishing more content on your own domain, is frequently the wrong move.

  • Sampled answers, annotated
  • Source influence by domain
  • Client, competitor and third-party split

The mechanism

Where an answer actually comes from.

Four steps, none of them mysterious. Most of the confusion about AI search disappears once an agency can see this sequence written down.

01The questionSomeone asks in their own words. There is no keyword, and no two people phrase it identically.
02RetrievalThe engine gathers candidate sources: pages it can reach, trusts and considers relevant to that phrasing.
03AssemblyIt writes one answer from what it gathered, naming whichever companies those sources named.
04AttributionSome engines link the sources they used. Where they do, the pages that decided the answer are observable.

Citation intelligence

Not just whether AI mentioned your client. Why it answered that way.

An answer is assembled from sources. Knowing which pages the model drew on is the difference between guessing at a fix and briefing one.

Sampled answerChatGPT · 09:14

Prompt: “What is the best accounting software for small businesses?”

For a small business with fewer than 25 employees, Acme Accounting is frequently recommended for its bank reconciliation and payroll add-on1. Reviewers note that Ledgerly has a lower entry price but limited multi-entity support2, while Countway is generally positioned at businesses with in-house finance teams3.

Client named
Yes
Position
#1 of 3
Cited as source
Yes
  • 1acmeaccounting.example/payroll-integration
  • 2accountingreview.example/ledgerly-review
  • 3countway.example/who-we-serve
  • Client
  • Competitor
  • Cited source
Sources influencing this categoryDemo data
Sources cited across sampled AI answers, with share of citations and change. Demonstration data.
SourceTypeCitationsShareChange
acmeaccounting.exampleClient3418%+12%
ledgerly.exampleRival2714%+4%
accountingreview.exampleReview1910%+21%
smallbizforum.exampleForum168%−6%
financeweekly.examplePress147%+9%
countway.exampleRival116%−2%

When a review site outranks your client’s own domain as a source, the fix is rarely another blog post.

The response

Three findings, three different pieces of work.

Citation data is only useful because the three cases it separates call for genuinely different responses, and two of them are not content production.

01

The source is yours

The client's own page is already cited. The work is to defend it: keep it current, keep it specific, and make sure the claims a model would quote are still the accurate ones.

02

The source belongs to someone else

A review site or an editorial page is carrying the category. Publishing a competing blog post does not touch that. Getting the client accurately represented on the cited source does.

03

There is no source at all

Nothing on the client's site answers the question in a form anything could quote. This is the most common finding, and usually the cheapest to fix.

See every client’s AI visibility in one place.

Start measuring where your clients appear, and what to do about it.