Citation share is the new keyword ranking.

Citation share is simple: of the prompts a buyer might use to research your category, what percentage of the AI-generated answers include your brand? You define the prompt set, run it across the engines you care about, and count how often your brand appears. That percentage (cited in X% of category-relevant prompts) is your citation share.

Compare that to what a keyword ranking actually measures: the position of a specific URL in a specific SERP at a moment in time. Keyword rankings capture one channel (Google organic), one format (link-list results), and are becoming less relevant as AI Overviews absorb more queries before a user ever sees a blue link. They don't tell you anything about what a buyer sees when they ask ChatGPT which security tools are worth evaluating.

Citation share measures something closer to the question that actually matters: when a buyer is forming their consideration set using an AI assistant, are you in it?

Keyword rankings require translation. "We moved from position 4 to position 2 for this term" means nothing to a CFO or a CEO without several intermediary steps: click-through rate curves, traffic estimates, conversion assumptions. The chain of inference is long and each link is an opportunity for scepticism.

"We're cited in 34% of ChatGPT prompts for our category, up from 18% three months ago" requires almost no translation. The person asking the question is a buyer. The AI is answering it. We're in or we're not. That's a much shorter inference chain to business outcome, and it maps directly to the buyer behaviour the leadership team is already hearing about anecdotally: "our customers tell us they researched us in ChatGPT before reaching out."

Citation share also travels across channels in a way rankings don't. A single citation-share number can aggregate across ChatGPT, Perplexity, Google AI Overviews, and Claude simultaneously. It becomes a single headline metric that captures visibility across the surfaces where buyers actually research, rather than requiring separate reports per platform.

You do not need a specialist tool to start. Here is the minimum viable setup:

1. Build a prompt set. Write 50–80 prompts that represent how a buyer in your category researches solutions. Include: "what are the best tools for X", "compare X vs Y", "how do teams typically solve X", "what should I look for in an X solution". Use real language, not SEO-formatted queries. This takes two to three hours the first time.
2. Run the prompts. Run each prompt against the three to five engines you care about most. Log every brand mentioned in each response, not just your own. This gives you competitive context as well as your own number.
3. Score your citation share. Count the prompts where your brand appeared, divided by total prompts. Do this per engine and as an aggregate. Record the date.
4. Track the trend. Repeat monthly, using the same prompt set. The absolute number matters less than the direction over time.

The prompt set is the most important thing to get right. A set of 50 poorly-chosen prompts will give you a misleading citation share. The prompts should represent real buyer intent. They should be based on how buyers in your category actually talk, not how you wish they talked. Talk to sales, read support tickets, look at what people type into your site search.

For the measurement framework to interpret these numbers against business outcomes (including the attribution problem when AI answers don't pass UTM parameters), I cover that in detail in how to measure AI search ROI when there's no click to track.

Citation share responds to entity coherence and source-worthiness more than to any on-page optimisation. The levers that move it are: getting your Wikidata entry correct, earning editorial coverage in publications models weight highly, building community-level recognition in the spaces where your buyers talk, and producing the kind of named-author original content that models treat as a citable source rather than as marketing copy.

What doesn't move citation share much: publishing more blog posts without improving the above, tweaking schema markup on pages that are already technically clean, adding more keywords to existing pages. These interventions are not without value. They matter for traditional organic and for Google AI Overviews specifically, but they have minimal effect on the citation rate you'll see in ChatGPT or Perplexity.

The practical priority order: reference sources first (Wikidata, Wikipedia), then editorial coverage, then structured authorship on your best content. The post on what source-worthiness means for SEO teams covers the full framework.