Citation Share: The KPI Behind GEO

The metric in one line

Citation Share is the share of AI-generated answers across a defined prompt set in which a brand appears.

It's the GEO equivalent of market share. The engine is the market. The answer is the shelf. The citation is the shelf placement.

Why a new KPI

SEO ran on traffic. Pageviews, sessions, conversions from organic search. Every analytics tool measured it.

GEO breaks that model. When the engine answers the question, the buyer doesn't click. There's no session to count. The brand was either named — or it wasn't.

Traffic is no longer the floor metric. Citation is.

The formula

Citation Share scores across five components:

How each component is scored

Citation Frequency — 40%

Take the prompt set — 50 to 100 buyer questions in the brand's category. Run each prompt through each of the five engines. Count the number of answers in which the brand appears. Divide by total prompts.

A brand that appears in 30 of 50 prompts has a Citation Frequency of 60%. Score weighted at 40% of total.

Cross-Engine Breadth — 20%

Five engines: ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews. A brand cited in all five gets full marks. A brand cited in one gets 20% of full marks.

Breadth matters because buyer segments use different engines. A brand cited only in ChatGPT is invisible to the rising Perplexity user base.

Query-Type Breadth — 20%

Different queries pull different brands. The four query types we track:

• Definitional ("what is X")
• Comparative ("X vs Y" or "best X for Y")
• Procedural ("how do I do X")
• Recommendation ("which X should I buy")

A brand cited only in definitional queries is half-built. The buyer's purchase question is usually recommendation or comparative.

Extractability — 15%

Even when an engine knows a brand exists, it can't cite it if its pages don't render as clean answer blocks. Extractability scores the brand's own properties: schema implementation, definition blocks, comparison tables, FAQ structure, header discipline.

This is the lever the brand controls directly. Most programs leave 30–50% of Extractability on the table.

Crawl Access — 5%

The basement floor. Can engines access the brand's site at all? robots.txt configuration, llms.txt presence, JavaScript rendering, crawl budget. Small weight, but a zero here zeros out everything above.

What good looks like

Benchmarks vary by category. In high-velocity consumer categories (beauty, hospitality, B2B software), a competitive brand needs Citation Share above 30% to show up reliably in buyer answers.

A brand below 10% is functionally invisible. A brand above 50% is the category default. Most established brands are sitting at 5–20% with no measurement in place.

How often to measure

Baseline once. Re-measure monthly. The compounding shows up in months 3–4 — that's when retrieval architecture and entity authority start reinforcing each other inside the engines.

Don't measure weekly. The signal-to-noise ratio is too low. Don't measure quarterly. By the time you find the gap, a competitor has already filled it.

What to do with the score

1. Baseline reveals the largest weighted gap.
2. Citation Frequency low? Build more retrieval anchors and earn more third-party mentions.
3. Cross-Engine Breadth low? Diagnose which engines aren't citing you and engineer for their architecture.
4. Query-Type Breadth low? Build the missing anchor types.
5. Extractability low? Fix schema, structure, and on-page architecture.
6. Crawl Access low? Fix robots, llms.txt, JS rendering immediately.

The bottom line

What you don't measure, you can't move. Citation Share is the only KPI that tells you whether the engine knows you exist — and whether the buyers using it will hear your name.