AI citation share is the percentage of AI-generated answers on a set of relevant prompts that cite a given brand or domain — the AI-era analogue of keyword rankings.
Because answer engines return prose instead of ranked lists, classic rank tracking doesn't apply. Citation share replaces it: define a panel of prompts your customers actually ask, run them on a schedule across ChatGPT, Perplexity, and Gemini, and record which domains get cited. Your share of those citations — and its trend — is the KPI.
Citation share responds to the same inputs as AI visibility: crawler access, extractable content, schema, and authority. It typically lags fixes by weeks, since engines refresh their retrieval indexes on their own cadence.
How to calculate it
Citation share is the number of answers that cite your domain divided by the total number of answers you sampled, usually expressed as a percentage. If you ask 50 prompts and your domain is cited in 8 of the answers, your citation share is 16 percent.
Write a panel of prompts your real customers would ask, including some that do not name your brand.
Run each prompt on each engine you care about, on a regular schedule.
Record which domains each answer cites or mentions.
Divide answers citing you by total answers, per engine and overall, and track the trend.
Why single results mislead
AI answers are not deterministic. The same prompt can return different sources on different runs, and results differ by engine, location and sometimes by time of day. A single answer that cites you, or does not, tells you very little. That is why citation share is measured across many prompts and repeated runs, and why you should watch the trend rather than any one reading.
What moves it
Whether the engine's crawlers can reach your pages at all.
Whether your content states the answer plainly and in extractable form.
Whether other credible sources mention and link to you.
How closely your pages match the questions in the prompt panel.
Frequently asked questions
How many prompts do I need?
Enough that one odd answer does not swing the result. Dozens of prompts, repeated over time, give a far steadier picture than a handful. There is no magic number; more prompts and more runs reduce noise.
Is citation share the same as share of voice?
They are related. Share of voice usually measures visibility across a set of keywords in classic search. Citation share applies the same idea to AI answers, counting how often you are cited rather than where you rank.
Can I track citation share for free?
You can sample prompts by hand to get a rough feel. Doing it consistently across engines and over time is easier with a tracker; Website Verdict's paid plans include one.