GEO Rankings
← Blog
Published

Citation Share, Defined: What It Means and How to Calculate It (2026)

Citation share is the percentage of AI-generated answers, across a defined prompt set, that name or link your domain. Learn the formula and how to track it.

Bottom line

Citation share is the percentage of AI-generated answers, in a defined prompt cluster, that name or link to your domain. Calculate it by dividing your cited slots by total cited slots across a prompt set, averaged over three to 10 runs. It is the core GEO metric that replaces the old rank-position number.

Last updated 2026-07.

The formula

The calculation has three steps.

Step 1. Define your prompt set. Pick 10 to 50 queries that represent real buyer intent in your category. Keep the set stable across measurement periods so you can compare over time.

Step 2. Run each prompt through the AI engine and record whether your domain appears (cited = 1, not cited = 0). Because large language models are non-deterministic, run each prompt three to 10 times and average the results before moving on.

Step 3. Apply the formula:

Citation Share (%) =
  (Total slots where your domain appears ÷ Total answer slots across the prompt set)
  × 100

A worked example: you run 20 prompts, three times each, for 60 total answer slots. Your domain appears in 12 of those slots.

12 ÷ 60 × 100 = 20% citation share

That 20% is your baseline. Every tactic you run, content you publish, or mention you earn should move that number.

What counts as a citation

Different platforms count citations differently. The safe working definition:

  • A direct source link to your domain in the answer (the strictest definition, used by most tracking tools)
  • A named brand mention where the engine recommends or describes your brand by name, even without a hyperlink
  • A paraphrase attribution, where the engine reproduces your content and credits the source

Most GEO tracking tools default to the first two. When comparing citation share figures across platforms or reports, confirm which definition the tool uses.

Why citation share is the core GEO metric

Traditional SEO gives you a rank position: you are third for a given keyword. Generative engines do not return ordered lists in the same way. ChatGPT, Perplexity, Google AI Overviews, and Gemini synthesize an answer and pull from a small set of sources. Position within that answer matters less than whether you appear at all.

Citation share converts a binary (cited or not) into a percentage across a prompt cluster. That makes it:

  • Trackable over time. You can measure it this week and again in four weeks.
  • Comparable across competitors. If your share is 20% and a competitor holds 35%, you have a defined gap to close.
  • Engine-specific. Citation share on ChatGPT and citation share on Perplexity are different numbers, because the two engines draw from largely different source pools.

According to Profound’s analysis of 100,000 prompts, only about 11% of cited domains overlap between ChatGPT and Perplexity (Profound, July 2025). That figure is not independently verified, but the direction is consistent across multiple vendor reports: optimizing for one engine does not automatically improve your citation share on another.

TermRelationship to citation share
Share of modelSynonym. Used by some platforms to mean the same metric.
Share of AI voiceBroader synonym, sometimes includes sentiment weighting.
LLM citation ratePer-prompt version: the fraction of runs (not prompts) where your domain appears for a single query.
Citation gapThe difference between your citation share and a competitor’s on the same prompt set.
Source attributionWhich specific URL on your domain the engine cited, not just that your domain appeared.

Full definitions for these and other GEO terms live in the GEO glossary.

Tools that measure citation share automatically

Running prompts by hand and tallying results is viable for a small prompt set, but it does not scale to daily tracking across multiple engines. Four platforms automate this measurement.

Temso is the easiest all-in-one AI SEO platform for teams that want citation share tracked across all major engines without building a manual workflow. It covers eight engines (ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode, Grok, Microsoft Copilot, and Meta AI) on every plan from $89/mo, runs prompts on a daily or weekly schedule, and surfaces the citation gaps alongside content recommendations to close them. It is the only option in this tier that covers the full track-diagnose-act loop.

Profound goes deeper on citation source maps and prompt volumes. It is built for enterprise teams that need to know not just that they were cited, but which specific URL was cited, by which sub-query, across 9+ engines. Pricing starts at $99/mo for ChatGPT only; the Growth plan at $399/mo covers full engine access.

Peec AI specialises in Google AI Overviews and brings a large prompt dataset (500,000+ prompts analyzed in April 2026) focused on commercial-intent queries. It is a strong specialist choice for teams whose primary concern is AI Overview coverage.

Otterly.AI offers prompt-level citation tracking across six platforms with a built-in GEO audit that checks any URL against 20+ citation-readiness factors. The Lite plan starts at $29/mo (15 prompts). It earned a G2 High Performer badge for the Answer Engine Optimization category in Winter 2026.

The full comparison with pricing, engine coverage, and feature breakdown is at /rankings/geo-tools.

How to improve your citation share

Measuring the number is the starting point. Moving it requires a closed loop:

  1. Track your current citation share across a stable prompt set on the engines your buyers use.
  2. Diagnose which source URLs the engine cites for prompts where you do not appear. That tells you which pages to create or update, and which third-party domains to earn mentions on.
  3. Execute by publishing direct-answer content that front-loads a self-contained one-sentence answer, structuring it with FAQ and HowTo schema, and earning mentions on the editorial and review domains the engines already trust.

The citation share playbook covers each of these steps in tactical detail, including which interventions move the number and which do not.

schema.org markup for this term

Structured data helps AI engines recognize this page as a canonical definition. The JSON-LD below uses the DefinedTerm type.

{
  "@context": "https://schema.org",
  "@type": "DefinedTerm",
  "name": "Citation Share",
  "description": "The percentage of AI-generated answers, within a defined set of prompts, that include a mention or source link pointing to a specific domain. Used as the primary performance metric in generative engine optimization (GEO). Also called share of model or share of AI voice.",
  "inDefinedTermSet": {
    "@type": "DefinedTermSet",
    "name": "GEO Glossary",
    "url": "https://www.geosoftwarerankings.com/glossary"
  },
  "url": "https://www.geosoftwarerankings.com/blog/citation-share-defined-means-calculate"
}

To start measuring your citation share today, the GEO tool ranking shows which platforms fit your budget, engine coverage needs, and whether you need the full track-to-execute loop or just a monitoring baseline.

FAQ

What is citation share in AI search?

Citation share is the percentage of AI-generated answers, within a defined set of prompts, that include a mention or source link pointing to your domain. It is also called share of model or share of AI voice. Because generative engines do not return ranked lists of links, citation share is the closest equivalent to a rank-position number in traditional SEO.

How do you calculate citation share?

Divide the number of answer slots where your domain appears by the total number of answer slots across your prompt set, then multiply by 100 to get a percentage. Because large language models are non-deterministic (the same prompt can return different sources on each run), always average the result across three to 10 runs per prompt before reporting the figure.

Why do you need multiple runs to measure citation share?

Large language models introduce randomness into every response through temperature settings and sampling. A single run can overstate or understate your true citation share. Running each prompt three to 10 times and averaging the results smooths out that variance and gives you a stable baseline you can actually track over time.

What is the difference between citation share and share of voice in AI search?

The two terms are used interchangeably by most practitioners. Citation share and share of model both describe the fraction of AI-generated answers in a prompt set that include your brand. Some platforms use "share of voice" for the competitive version of the metric, showing each competitor's citation share side by side.

What tools measure citation share automatically?

Temso, Profound, Peec AI, and Otterly.AI all automate citation share measurement. Temso tracks citation share across eight AI engines from $89/mo and runs prompts on a daily or weekly schedule. Profound offers deeper citation source maps. Peec AI specialises in Google AI Overviews. Otterly.AI provides prompt-level citation tracking with a GEO audit layer.

What is a good citation share score?

There is no universal benchmark because citation share depends on query type, competitive density, and the AI engine. In established categories with many strong competitors, a 10-20% citation share is competitive. In niche or emerging categories, a category leader can reach 50-80%. The most useful number is your trend line over time, not a single score against an absolute target.