AI Share of Voice (AI SOV)

Why It Matters #

Machine Relations uses AI Share of Voice to compare observed brand presence within a declared query-and-engine panel. It is a descriptive measurement, not proof of buyer behavior, revenue, or the cause of a change in visibility.

The Denominator Decides the Number #

Three different calculations are reported under this one name, and only one of them is a share. Which denominator a tool uses changes the answer by an order of magnitude, so the denominator must be declared with every value.

Denominator	What it actually measures	Why it misleads
Total prompts tracked	Visibility rate: how often the brand appears at all	Not a share. A brand named alongside nine competitors scores the same as a solo mention
All brand mentions in a pre-set competitor list	Share inside a configured pool	A brand rising in the answers is invisible until someone adds it to the list
All brand mentions observed in the same answers	Competitive share as the engines represent the category	None of the above; this is the MR contract

Machine Relations definition: only the third is AI Share of Voice. A brand appearing in 30 of 100 prompts does not hold 30% share; if each of those answers names six brands, its share of the observed brand conversation is nearer 5%. Report visibility rate under its own name when that is what was counted.

AI Share of Voice vs. Share of Citation #

These two metrics answer different questions and use different measurement contracts.

Share of Citation, defined at /glossary/share-of-citation, is a distinct citation-source presence and depth metric under its own declared contract. It tracks when an answer engine cites a source domain or source asset; it is not an AI SOV denominator and should not be redefined here.

Metric	Measures	Frame	Example
AI Share of Voice	Brand X has Y% of all qualifying brand mentions in the panel	Relative (vs. competitors)	"We hold 35% of AI brand mentions; competitor holds 22%"
Share of Citation	Citation-source presence or depth under the declared Share of Citation contract	Scoped citation metric	"Our source domain is cited in 28% of observed answer runs for this segment"

Report the two observations separately. AI Share of Voice records competitive mention breadth; Share of Citation records citation-source behavior. They can inform the same visibility diagnosis, but one does not substitute for the other.

Measurement Framework #

Step	Action
1. Define query set	Category-relevant queries a buyer would actually ask
2. Run across engines	ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews
3. Record mentions	Which brands appear in each response, per engine
4. Normalize entities	Count a brand and its products as one entity. "HubSpot" and "HubSpot CRM" in one answer are one brand, not two; without this the denominator fragments and the share drifts
5. Calculate per-brand	Brand mentions / Total brand mentions = AI SOV per engine
6. Aggregate or segment	Report as cross-engine average or per-engine breakdown, with the engine set and weighting disclosed

Worked example (arithmetic illustration, not an observed benchmark): a brand with 45 mentions out of 200 total brand mentions holds 22.5% AI Share of Voice.

AI answer engines are non-deterministic: identical queries can produce different responses and cite different sources. Repeated-sampling research treats citation visibility as an estimate of an underlying response distribution and finds that single-run measurements can imply more precision than the observations support (Sielinski, arXiv 2603.08924, 2026).

Platform-Reported Share of Voice Is a Different Panel #

On 27 May 2026 Google announced AI performance insights in Merchant Center, including a share-of-voice report benchmarking a brand's visibility on AI Mode, AI Overviews and the Gemini app against similar brands, with product-term share of voice and discovery, evaluation and purchase stages; it was described as rolling out to the United States, Canada, Australia, India and New Zealand (Google Merchant Center Help, 27 May 2026).

A platform's own share-of-voice figure is scoped to that platform's surfaces, its product corpus and its undisclosed comparison set. It is one engine's panel, not a category-level measurement, and it is not comparable with a cross-engine AI SOV value. Report the two separately and name the panel behind each.

What It Is Not #

• Not traditional Share of Voice, which counts media mentions across press, social and broadcast. AI SOV counts brand mentions inside AI-generated answers.
• Not search share of voice, which measures ranking visibility across keyword SERPs. AI SOV measures mention breadth in synthesized answers that often bypass those results entirely.
• Not Share of Citation, which is the cited-source layer described above. Keep mentions, cited sources, recommendations and positions as separate labels unless a protocol explicitly blends them.

Treat a single-run snapshot as one observation, not a stable benchmark. Record the query set, engine, collection time, and sampling method with every reported value.

Common Mistakes #

1. Reporting a visibility rate as a share. Name the denominator with the number.
2. Unlabeled engine scope. A result from one engine describes only that engine; name the observed surface.
3. Changing query sets silently. Keep the query set fixed for comparisons or disclose the change as a new measurement basis.
4. No competitive baseline. Tracking AI SOV in isolation misses the point. The value is in competitive comparison.

---

FAQ #

How is AI Share of Voice different from traditional Share of Voice? Traditional Share of Voice measures media mention volume across press and social channels. AI Share of Voice measures competitive brand-mention share inside AI-generated answers. The data sources, measurement methods, and strategic implications are structurally different — a brand can dominate traditional SOV and be invisible in AI responses.

How often should AI SOV be tracked? Choose a collection cadence before measurement and keep the query set and collection protocol stable across comparable cycles. A protocol change creates a new measurement basis and should be disclosed.

What tools measure AI Share of Voice? AI Share of Voice can be calculated from logged answer observations that identify the engine, query, observed brands, and collection time. The calculation does not require a specific vendor.

Can AI Share of Voice be gamed? AI Share of Voice records observed presence; it does not establish why that presence changed. Any explanation of a change requires separate evidence.

How is AI Share of Voice different from AI visibility? AI visibility is the broader question of whether a brand appears in AI-generated answers at all. AI Share of Voice is the competitive refinement of it: of all the brand mentions in those answers, how many are yours. Visibility asks whether you are present; share of voice asks how much of the observed conversation you hold.

What is a good AI Share of Voice percentage? There is no universal benchmark, because the value is set by competitive density. In a panel whose answers name five brands, mathematical parity is 20%; at ten brands it is 10%. Compare a brand against its own panel's parity point and its own prior cycles, never against a figure from another category.