Share of Voice in AI Search Explained

Full definition

Share of Citation (SoC) is the percentage of AI-generated responses, within a fixed set of queries, in which a specific brand or domain appears as a cited source. The formula is:

SoC = (responses citing your domain / total responses sampled) x 100

The term was coined by AuthorityTech in 2024 and has established itself as the primary visibility metric in generative search, displacing traditional Share of Voice, which counted media mentions and organic positions.

The distinction from classic Share of Voice is structural. SoV measured exposure in environments where visibility could be distributed across many brands simultaneously. Generative engines, by contrast, synthesize a single response and cite between three and eight sources per query. According to an analysis of 366,087 citations in AI search conversations (Yang et al., arXiv:2507.05301, 2025), citation distribution is extremely concentrated within a small group of domains. If an engine cites you, it almost certainly does not cite a competitor in that same response.

A competitive variant known as AI Share of Voice (AI SoV) measures relative share among brands: of all brand citations in the query set, what percentage belongs to each brand. It is, by definition, a zero-sum metric.

Why it matters in 2026

Research cited by AuthorityTech (2026) indicates that 80% of the sources cited by AI platforms do not appear in the Google top 10 for the same query. Moz published in 2026 that 88% of citations in Google AI Mode do not overlap with the organic top 10. Leadership in classic SEO and leadership in SoC are, in practice, independent measurements.

For home services and construction companies, where purchase decisions are complex and comparative, this disconnect is critical: a business can dominate organic rankings and be entirely invisible when a buyer asks an AI engine which company to hire.

SparkToro and CXL estimate that 69% of queries in AI engines end with no click to any website. This means the only brand footprint left in those sessions is the citation inside the generated response. If you are not cited, you do not exist for that user in that moment.

How it works

Measuring Share of Citation follows a repeatable process:

1. Define a fixed set of queries representative of the sector — a minimum of 30 to 50 real buyer prompts.
2. Run each query in the relevant engines: ChatGPT, Perplexity, Gemini, Claude, Google AI Mode.
3. Record in which responses the domain appears as an explicit citation with a URL.
4. Calculate SoC = domain citations divided by total responses, multiplied by 100.
5. Repeat the process at regular intervals to detect trends.

A single measurement is unreliable: AI engines are non-deterministic and may return different sources for the same query across sessions. To obtain confidence intervals below five percentage points, repeated sampling with at least three runs per query is recommended.

Difference from related metrics

Metric	What it measures	Main limitation
Classic Share of Voice	Media mentions and SERP positions	Does not predict AI source selection
AI Share of Voice	Brand citation share among competitors	Includes unlinked mentions
Share of Citation	Explicit URL citations in AI responses	Requires active sampling
Entity Resolution Rate	Whether the engine correctly identifies the brand	Does not measure whether it selects it as a source

The difference between AI SoV and SoC is operationally significant: the former counts any brand mention, whether or not it includes a linked URL. The latter counts only citations with an associated URL. Because only linked citations generate referral traffic from AI engines, SoC is the indicator most directly tied to business outcomes.

Related terms

Citability, Fan-out query, RAG (Retrieval-Augmented Generation).