Citation Share is the percentage of AI engine answers to a defined set of category buyer prompts that name your brand. It is the engine-era successor to market share and share of voice. Those metrics counted purchases and press clips, while Citation Share counts the one surface that now sits between a buyer and a decision: the composed answer.
Pillar hub. Updated October 2026.
What does Citation Share measure?
Citation Share measures how often a brand is named inside AI-generated answers to the questions buyers actually ask. Every category has a finite set of those questions. Best crisis PR firm for a public company. Which beauty brands are dermatologist-recommended. Who are the top hospitality PR agencies. ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews answer them with a short list of named entities, and Citation Share is your share of that list, measured across the prompt set, across engines, over time.
Three properties make it different from the metrics it replaces:
- It is winner-concentrated. A composed answer names three to seven entities and there is no page two. Rank eleven is invisible in a way that rank eleven in search never was.
- It is durable. Engine portraits are built from a corpus that accumulates over years. A quarter of heavy coverage does not move it, while a decade of consistent named-principal material does.
- It is measurable without the platform's cooperation. You do not need engine analytics. You need a prompt set, a sampling cadence and disciplined logging.
How is Citation Share different from market share and share of voice?
Citation Share differs from market share and share of voice in what it counts and where it is measured. Market share counts purchases, share of voice counts media mentions, and Citation Share counts named appearances inside answers. A brand can lead on the first two and still be absent from the third.
| Market share | Share of voice | Citation Share | |
|---|---|---|---|
| What it counts | Purchases or revenue | Media mentions | Answers that name the brand |
| Where it is measured | The market | The press | AI engines |
| What moves it | Distribution and price | Coverage volume | Source diversity, entity consistency, retrievable structure |
How do you calculate Citation Share?
You calculate Citation Share by dividing the answers that name your brand by the total answers you sampled. The work is in building a fair sample, in four steps.
- Define the prompt set. Use 50 to 250 prompts that mirror real buyer language across the funnel: category discovery, shortlist, comparison, objection and post-decision validation.
- Sample across engines. Run each prompt on every engine that matters to your buyers, on a fixed cadence. Monthly is the minimum useful interval.
- Score each answer. Record named or not named. Then, separately, record framing: leading, listed or adverse.
- Compute the share. Citation Share equals answers naming you divided by total answers sampled. Track it by engine, by funnel stage, and against the three competitors who appear most often.
The second-order metrics matter as much as the headline number. Source diversity is how many distinct domains support your citations. Entity consistency is whether the engine resolves your brand to one entity or three. Temporal depth is how far back the retrievable corpus runs.
What did the CPG Citation Index find?
The CPG Citation Index 2026 found that scale and brand heritage did not predict who gets cited. The study audited 75 consumer product queries across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, for 375 responses in total, and asked which brands appear inside generative AI search. The figures below are the study's own, reported as published.
Why is a $50 billion parent invisible in AI answers?
Unilever owns about $50 billion in revenue, yet its brands scored very differently. Vaseline appeared in 8 percent of skincare queries, Lipton in 12 percent of tea queries, Hellmann's in 8 percent and Suavitel in 0 percent. Dove appeared in 68 percent. They share a parent and resources, and the difference the study points to is positioning. Dove markets as clinical, with dermatology and sensitive-skin language, while Vaseline markets as a heritage cosmetic. Models trained on clinical literature cite the first and largely skip the second.
What did the study say wins citations?
- Clinical positioning beats scale. Neutrogena (85 percent), CeraVe (80 percent) and La Roche-Posay (78 percent) lead skincare on dermatology credentials, while Aveeno, backed by a much larger parent, appeared in 45 percent.
- Portfolio breadth creates a moat. L'Oréal holds Maybelline (80 percent, mass), L'Oréal (75 percent, mid-market) and Urban Decay (70 percent, prestige), so it is visible at three price tiers.
- Community infrastructure beats heritage. Celsius (75 percent) outranked Gatorade (60 percent) in fitness queries, which the study attributes to Reddit sponsorships, creator deals and Amazon reviews rather than product superiority.
- Review volume works as discovery infrastructure. The study reports that food brands citing Amazon appeared 38 percent more often in recommendations, because review volume, ratings and Q&A responses feed the models.
The practical reading is that retail shelf space, TV reach and heritage do not carry into the answer layer. Discovery mechanism is what counts there. The study also announced five further category indexes through Q4 2026: beauty, food, home care, health and wellness, and personal care.
Why do brands with strong PR still score low?
Brands with strong PR still score low because the causes are structural, not editorial. The most common diagnostic finding is a brand with real earned media and near-zero Citation Share, and it usually traces to one of four problems.
- Concentrated coverage. Forty placements in one outlet family read to an engine as one source, not forty.
- Entity fragmentation. Inconsistent naming, no canonical about page and no schema mean the engine never consolidates the portrait.
- Unretrievable formats. Wire releases, PDFs, gated assets and image-only announcements contribute almost nothing to the corpus.
- No named principal. Engines retrieve people faster than they retrieve logos, so a brand with no attributable human voice gives them nothing to anchor to.
How do you improve Citation Share?
You improve Citation Share by fixing structure first and then building a diversified, attributable corpus over several quarters. Entity and structural fixes can register within weeks, while corpus-driven movement is slower.
- Fix the entity. Use one canonical name, one retrievable about page and consistent schema everywhere the brand appears.
- Diversify the sources. Earn coverage across many independent outlets, not a deep pile in one family.
- Publish in retrievable formats. Plain, indexable pages with answer-first structure beat PDFs and image announcements. The answer engine optimization guide covers the writing pattern.
- Put a named person behind the claims. Dated, signed, first-person material gives engines something to anchor to.
- Re-sample on a schedule. Rerun the same prompt set monthly so movement is attributable to specific changes.
Meaningful share change generally follows two to four quarters of sustained, diversified, named-principal publishing. For the execution layer, see the GEO for PR checklist.
The library
- AI Visibility Research Library: every index and finding on this site, sorted by industry.
- How AI Engines Decide What to Cite About a Brand: the five weighting factors.
- 15 Websites Now Control 68% of Every AI Answer: the source-concentration map.
- The Press Release Retrieval Study: why 81% of wire releases produce zero citations.
- Gartner Just Said What We've Been Building: third-party validation of the thesis.
- Two Brands Own the Home: category-level Citation Share concentration.
- Golf Equipment Brands Are Losing Sales They Never See: the commercial cost of low share.
- GEO for PR: A Practical Checklist: the execution layer.
- How to Rank in ChatGPT Search: engine-specific tactics.
Frequently asked questions
What is Citation Share?
Citation Share is the percentage of AI engine answers to a defined set of category buyer prompts that name your brand. It measures presence inside composed answers rather than clicks, impressions or press volume.
How is Citation Share different from share of voice?
Share of voice counts your proportion of media mentions. Citation Share counts your proportion of answers. A brand can hold high share of voice in a single outlet family and still be absent from every engine answer in its category, because engines weight source diversity, entity consistency and retrievable structure rather than raw mention volume.
How do you measure Citation Share?
Build a prompt set that mirrors real buyer questions, run it across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews on a fixed cadence, score each answer for whether the brand is named and how it is framed, then divide named answers by total answers sampled.
How long does it take to move Citation Share?
Entity and structural fixes, such as canonical naming, schema and a retrievable about page, can register within weeks. Corpus-driven movement is slower, and meaningful change generally follows two to four quarters of sustained, diversified, named-principal publishing.
Does a big company automatically have a high Citation Share?
No. The CPG Citation Index found brands under the same parent with very different scores, such as Dove at 68 percent and Suavitel at 0 percent. Positioning and corpus, not company size, drove the difference.