What is citation share of voice?
Citation share of voice is the share of all AI citations, in a defined set of AI answers, that go to your website. If 50 AI answers about project management software cite 300 sources between them, and 30 of those citations point to your domain, your citation share of voice is 10%.
The metric answers a competitive question an individual AI citation count cannot: out of all the credit AI engines hand out for a topic, how much goes to you and how much goes to each competitor.
The term is new and has no published standard. AI visibility tools report "share of voice" figures [1], but they do not all count the same thing. This entry sets out a definition precise enough to compare one month with the next.
How do you calculate citation share of voice?
To calculate citation share of voice, divide the number of citations to your domain by the total number of citations across the same set of AI answers, then multiply by 100.
Citation share of voice = (citations to your domain ÷ all citations in the answer set) × 100A worked example on one engine, one prompt set and one week:
| Domain | Citations | Citation share of voice |
|---|---|---|
| yoursite.com | 30 | 10.0% |
| competitor-a.com | 54 | 18.0% |
| competitor-b.com | 21 | 7.0% |
| reddit.com | 45 | 15.0% |
| All other domains | 150 | 50.0% |
| Total | 300 | 100% |
Four decisions change the result, so record them every time:
- The prompt set. Which questions were asked, worded how. Use the same set each period.
- The engine. Google AI Overviews, AI Mode, ChatGPT, Perplexity and others cite different sources. Report each separately before combining.
- The denominator. Either every citation from every domain (a market view, as above) or only citations to a named competitor set (a rivalry view). The rivalry view always gives higher numbers. Say which one you used.
- Counting rule. Count each domain once per answer, or every link it gets. Search Insider recommends counting each domain once per answer, so a site linked three times in one answer does not triple its credit.
What is the difference between citation share of voice and share of voice?
Citation share of voice measures how much of the citation credit in AI answers goes to a brand, while traditional share of voice in SEO estimates how much of the organic search visibility for a keyword set a brand holds. SEO share of voice is usually weighted by search volume and click-through rate by ranking position. Citation share of voice has no reliable volume or click data to weight by, because AI engines publish neither [3].
The two often move together for Google's AI features. Search Insider found AI Overview citations "moved with the organic results rather than independently" [6]. A brand that loses organic share on a topic should expect its citation share on Google's AI features to fall with it.
What is the difference between citation share of voice and share of model?
Citation share of voice counts links to a website, while share of model counts how often a brand is named in AI answers, linked or not. They measure different things and can disagree sharply:
| Citation share of voice | Share of model | |
|---|---|---|
| Counts | Sources linked under the answer | Brand names inside the answer text |
| Credits | The website that was cited | The brand that was mentioned |
| A review site citing you | Credit to the review site | Credit to you |
| Best for | Content and SEO teams | Brand and PR teams |
A brand can have a high share of model and a low citation share. That happens when AI engines recommend the brand by name but cite review sites, forums or publishers as the source. In that case, the pages that shape the answer belong to someone else, which is worth knowing before deciding where to spend effort.
How many prompts do you need to measure citation share of voice?
No published research sets a minimum number of prompts for citation share of voice, so any figure quoted as a standard is a convention, not a finding. What is known is that the number moves for reasons unrelated to your site:
- AI answers change over time. Search Insider's AI Overview study notes "the same searches may cite different sources later" [5].
- Small edits to a question change the sources. Deleting a stop word from a search left 69% of AI Overview cited sources the same, and swapping a stop word for its opposite left 20% [6].
- Each answer cites a handful of sources. Search Insider recorded 5,834 citations across 947 AI Overviews, about six per answer [5], so a small prompt set produces a small total and large swings.
The practical rules that follow: use enough prompts that one answer cannot swing the result by several points, repeat each prompt more than once per period, and only treat a change as real if it persists across periods.
Can you get citation share of voice from Google Search Console?
No. Google Search Console cannot give citation share of voice, because it reports only your own site and only impressions. The generative AI report rolled out worldwide in 2026 shows how often your pages appeared in AI Overviews, AI Mode and Discover AI features, by page, country and date, with no clicks [3]. It does not show competitors' citations, which the denominator needs, and launched without query data [4].
Search Console is still useful alongside the metric. It shows whether total AI impressions are growing, which a sampled prompt set cannot.
Where did citation share of voice come from?
Citation share of voice adapts share of voice, a long-standing advertising and SEO metric, to AI answers. Researchers measured the visibility of individual sources inside generative answers in the GEO paper first published on November 16, 2023, which reported that optimization could raise a source's visibility by up to 40% [2]. Commercial AI visibility tools later added "share of voice" reporting for AI answers [1], without a shared definition.
Why does citation share of voice matter?
Citation share of voice matters because it is the only AEO metric that puts your AI visibility in competitive context. A raw citation count can rise while competitors rise faster. A share cannot.
It also exposes who actually owns a topic in AI answers. In many categories the largest single share goes not to any brand but to forums, review sites or publishers. Search Insider's AI Overview data includes 161 searches where a top-three page carrying the current year was passed over and the AI Overview led with an undated source instead [5]. Knowing which outside domains hold the share tells you where to earn coverage, which is often more effective than rewriting your own pages.
What are the limits of citation share of voice?
Citation share of voice measures visibility, not traffic or revenue. AI engines do not publish how often their citations are clicked, and Google says nothing extra is needed to appear in its AI features beyond normal SEO [7], so there is no documented lever that moves the metric directly.
Treat it as a trend line on a fixed method. Compare a brand with itself over time, on the same prompts and engines. Comparing two vendors' share of voice figures, or one company's figure with another's, is not meaningful unless both used the same prompts, engines, denominator and counting rule.
When should you use citation share of voice?
Use citation share of voice when you need to compare your AI visibility against named competitors on a topic that matters commercially, and report it monthly or quarterly on a fixed prompt set. Use a raw AI citation count for a single page or campaign, and share of model when the question is about brand recommendation rather than source credit.