Citation analysis

What the analysis looks for

Run across a prompt set, citation analysis usually reveals a pattern: a specific review site dominates a query intent, Reddit threads outrank brand PDPs for comparison questions, or a competitor's longer PDP wins because it answers one question directly that yours implies. Each pattern maps to different work — review outreach, community presence, or answer-coverage rewrites — and remeasurement a week or two later confirms whether the engine picked up the change.

The discipline matters because engines cite few sources. Perplexity's 3–7 citations per shopping answer mean a handful of winners and a long tail of invisible pages; knowing who the winners are and why is the entire game of displacing one.

Why it matters for ecommerce

Without citation analysis, visibility tooling reports a number without a cause. A team sees its cite rate fall and has no basis for what to change. With it, the same fall reads as a diagnosis: "competitor X published three articles covering this query intent; three of our PDPs lost slots to them." That is actionable the same week.

It also grounds strategy. If independent review sites dominate a category's citations, the highest-leverage work may be review coverage, not more PDP copy. If forums dominate, community presence beats schema work. Citation analysis is how a brand avoids spending a quarter optimizing the wrong surface.

A diagnosis in practice: an example

A supplements brand runs citation analysis on its flagship protein powder. Prompts like "best whey protein for post-workout under $40" return answers where five of seven cited sources are independent review sites and one is a Reddit thread; the brand's PDP is rarely cited (illustrative example). Diagnosis: reviewer authority beats brand authority for this intent. The targeted program: seed product with independent reviewers, aggregate verified reviews on the PDP with AggregateRating schema, and add a comparison FAQ answering the same questions the Reddit thread debates. Re-measurement six weeks later shows the PDP entering the citation set on two of the five engines.

How it relates to neighboring terms

Citation analysis explains movements in AI visibility and the per-SKU citation score; it consumes the raw stream that prompt tracking produces; and it borrows its name from the academic discipline of studying reference networks — the AI-search variant studies a model's retrieval choices instead of journal references. For how one engine selects and attributes sources, Perplexity's own published materials are a useful primary source.

How eCommerce Insights does it

Every tracked answer stores its full citation set, so each SKU's ledger shows not just whether it was cited but who was cited instead — grouped by domain type (review site, forum, competitor, retailer). The recommended diffs draw on that context: when review sites dominate an intent, the recommendation says so rather than proposing another copy rewrite.

Related terms

• Citation score — the per-SKU metric this analysis explains.
• Prompt tracking — the measurement loop that supplies the raw answers.
• AI visibility — the outcome metric citation analysis diagnoses.
• Brand mentions — the adjacent brand-level signal stream.
• Query fan-out — how engines expand one prompt into many retrievals.