Share of Citation is your qualifying citations as a percentage of all qualifying citations counted in the same sampled answers — your share of the cited-source pool across a defined set of buyer-intent queries. It answers a question Share of Voice does not ask: AI engines do not rank pages, they select sources, and a brand that appears in rankings but is never selected as a cited source does not appear in the answer. The two measures use different denominators and are not substitutes for each other.
Why Share of Voice Does Not Measure Source Selection #
Share of Voice measures how often a brand appears across a set of search results. That worked when buyers scrolled ten blue links. AI engines operate differently: they synthesize a single answer from a small number of selected sources, and the rest of the web does not appear in it.
An MIT-led study of 24,000 queries across 243 countries found that AI search surfaced significantly fewer long-tail information sources and lower response variety than traditional search in its 2024-2025 sample (Aral, Li and Zuo, arXiv, 2026). The implication is direct: fewer citation slots mean the difference between being a cited source and being absent is binary, not gradual.
The Citation Share Index — a cross-category analysis spanning 21 studies — confirms that revenue rank consistently differs from citation rank. The largest brands by revenue rarely hold the strongest positions in AI-generated answers. This gap represents the core measurement problem Share of Citation solves: traditional brand-awareness metrics do not predict whether AI engines will select your content as evidence.
The Share of Citation Formula #
Machine Relations coined the term and defines it against the citation pool, and that is the formula this analysis uses (Share of Citation, the canonical definition):
Share of Citation = (your qualifying citations ÷ all qualifying citations in the same sampled answers) × 100
The denominator is every cited-source slot the sample issued, not the number of answers. On the Machine Relations run of April 17, 2026, 30 tracked queries returned 547 cited-source slots and 12% of them resolved to AuthorityTech. That is arithmetic, not a benchmark.
Divide by the answers instead of the citations and you have computed a different metric from the same sample: citation rate, eligible answers containing at least one citation to you over all answers sampled. Both are measurable, neither substitutes for the other, and a figure published without its denominator is not a measurement — so declare which one a number uses before trending it or setting it beside a competitor's.
The arithmetic matters more than it looks. In SimilarWeb's worked example, 179 buyer-intent prompts were tracked through ChatGPT, 19 prompts generated citation events, and 3 of those answers cited the tracked brand (SimilarWeb, 2026). The approximately 16% figure is 3 ÷ 19 — an answer-level ratio over answers that produced a citation, which is a citation rate, not a share of the pool. Three defensible denominators sit inside that one example: all prompts tracked (3 of 179, 1.7%), prompts that produced any citation (3 of 19, 15.8%), and the total citation slots those answers awarded, which the example does not report and which is the only one of the three that yields a Share of Citation.
The critical distinction from AI Share of Voice: citation share measures trust signals — whether engines use your content as evidence — not recognition signals like brand mentions without source attribution.
What to track:
- Total prompts in the tracked set (buyer-intent, not vanity queries)
- All qualifying citations those answers issued — the denominator
- Citations attributed to your domain — the numerator
- Prompts that generate any citation event, kept separately as citation rate
- Per-engine share (engines diverge significantly)
Engine-Specific Citation Divergence #
Not all engines cite the same way. Research analyzing 602 controlled prompts across ChatGPT, Google AI Overview, and Perplexity — producing 21,143 valid citations and 23,745 citation-level feature records — establishes a two-stage framework for understanding the difference.
Citation selection is whether an engine retrieves and links to your page. Citation absorption is whether the engine actually uses your page's language, evidence, structure, or factual support in its generated answer. These are separate outcomes that require separate measurement.
The divergence matters for founders:
- Perplexity and Google cite more sources on average, spreading citation slots across a wider set of pages
- ChatGPT cites fewer sources but demonstrates substantially higher average citation influence per cited page — meaning each citation carries more absorption weight
A Share of Citation score measured only on one engine will misrepresent actual AI visibility. The Citation Share Index methodology tracks approximately 62 buyer-intent prompts across five engines (ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews) per category study, and the engine-level variance within each study is the most actionable data point for operators.
What Gets a Page Selected as a Source #
Measurement without a theory of selection is just counting. Research on 1,702 citations from 1,100 unique URLs across three AI engines using the GEO-16 auditing framework identifies three structural predictors of citation inclusion:
- Metadata and freshness — pages with accurate, current metadata and recent publication or update dates are selected at higher rates
- Semantic HTML — proper heading hierarchy, structured content blocks, and machine-readable formatting increase citation probability
- Structured data — schema markup and extractable evidence types (definitions, numerical facts, comparisons, procedural steps) correlate with selection
The citation absorption research adds that pages with greater absorption influence tend to feature longer content, structured formatting, semantic alignment with the query, and what the authors call "extractable evidence types" — direct answers, tables, named data points, and step-by-step procedures.
This aligns with the Machine Relations framework for content structure and AI citation rates: the structural properties of a page determine whether it gets selected, independent of domain authority or traffic volume.
Five Patterns from Cross-Category Citation Research #
The Citation Share Index examined approximately 28 entities per category across 21 industry verticals. Five patterns appeared in every category:
| Pattern | Finding | Founder implication |
|---|---|---|
| Native sources beat legacy incumbents | Category-specific publications cite more frequently than established authorities | Build depth in your vertical, not breadth across topics |
| Reddit and primary data dominate experience queries | Community discussion and original data carry greater weight for ownership/due-diligence questions | Publish original research and benchmarks, not derivative commentary |
| Named individuals exceed institutional citations | Founders and practitioners achieve higher citation share than parent organizations | Attach a named expert to your content, especially in crisis, defense, and healthcare |
| Revenue rank differs from citation rank | Largest brands by revenue rarely hold strongest AI citation positions | Market share does not transfer to AI visibility automatically |
| First-mover advantage persists | Brands establishing early authority retain positions through model updates | The cost of waiting increases as citation positions calcify |
The third pattern — named individuals exceeding institutional citations — has direct implications for how founders structure their content strategy. A founder publishing under their own name with clear expertise signals may achieve higher Share of Citation than the same content published under a corporate byline.
Building a Share of Citation Measurement System #
A minimum viable measurement system requires five components:
1. Define the prompt set. Select 50–100 buyer-intent queries that represent how your customers actually ask questions. Vanity queries ("what is [your company]") inflate scores without measuring competitive visibility. Use queries where a buyer is making a decision, comparing options, or seeking evidence.
2. Run across engines. Track ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews at minimum. Single-engine measurement misrepresents actual visibility because engines diverge in both citation selection and absorption behavior.
3. Count citation events, not mentions. A citation is a linked source attribution. A brand mention without a source link is recognition, not citation. The distinction between trust signals and recognition signals is what separates Share of Citation from Share of Voice.
4. Calculate per-engine and aggregate share. Report both. Aggregate share shows competitive position. Per-engine share reveals where to invest: a brand with strong Perplexity citation share but weak ChatGPT share has a structural gap worth diagnosing.
5. Track over time. Monthly measurement establishes baselines. The Citation Share Index finding that first-mover advantage persists through model updates means early measurement creates compounding strategic advantage — you see shifts before competitors recognize them.
How Share of Citation Connects to Source Architecture #
Share of Citation is not a content-marketing metric. It is a source-architecture metric — it measures whether your published assets are structurally eligible for selection by retrieval systems.
The Machine Relations Index tracks citation behavior across six AI engines and multiple verticals, using citation breadth (how many engines cite a source), query diversity (how many distinct queries trigger citation), and temporal consistency (how stable citations are over time) as components of source authority measurement. Share of Citation is the founder-accessible version of this measurement: it answers the question "of the sources AI engines did cite, how much of that space was mine?" without requiring infrastructure-level tracking.
The practical connection: every structural improvement that increases Share of Citation — clearer entity definitions, extractable evidence blocks, proper metadata, citation-ready content architecture — also compounds across engines and time. A page that gets cited by three engines rather than one has higher source authority, which feeds forward into future retrieval decisions.
FAQ #
What is a good Share of Citation score? #
There is no universal benchmark. The Citation Share Index shows that in most categories, citation concentration is high — a small number of sources capture the majority of citations. A 10–20% share in a competitive category indicates strong positioning. The relevant comparison is within your category and query set, not across industries.
How is Share of Citation different from AI Share of Voice? #
Share of Citation measures trust signals — whether AI engines use your content as a cited source. AI Share of Voice measures recognition signals — whether your brand is mentioned at all, with or without source attribution. A brand can have high share of voice (frequently mentioned) and zero share of citation (never treated as a source).
Which AI engines should I track? #
At minimum: ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. These five engines demonstrate significant citation divergence — a source cited heavily by Perplexity may be absent from ChatGPT responses, and vice versa. Single-engine measurement creates blind spots.
How often should I measure Share of Citation? #
Monthly for baseline tracking. Weekly during active content campaigns or after major site-structure changes. The first-mover persistence pattern means that citation positions are relatively stable between model updates but can shift when new high-quality sources enter a category.