Insight
Learn how to measure ai share of voice, then interpret 0%, 50%, and 100% without mistaking a sample score for market dominance.
AI share of voice measures your brand's share of category mentions relative to competitors across answer engines (HubSpot). To calculate it, divide your brand's AI mentions by total AI mentions across all brands in the declared category, then multiply by 100 (Semrush). The result describes only the observation frame you counted. It does not establish your share of all answers, citations, buyers, or market demand. Keep mentions, citations, and sources separate so each number retains a clear meaning. Compare scores only when the category and comparison frame are relevant and fixed. A higher or lower score records a change in the sample, but the change alone does not prove what caused it.
AI share of voice answers a relative question: how visible is one brand compared with the other brands in its market? Semrush defines the metric as brand visibility compared with competitors, based on AI mentions across the category (Semrush). HubSpot defines it as how frequently and prominently a brand appears in answer-engine responses relative to competitors (HubSpot).
That relative frame matters. A mention count says how many observed answers named a brand. AI share of voice places that count over the mentions observed for all brands in the defined category. The percentage therefore describes a share of category mentions, not an isolated total (Semrush).
HubSpot contrasts this metric with traditional share of voice. Traditional share of voice covers channels such as advertising, organic search, and social media. AI share of voice focuses on answer engines such as ChatGPT, Gemini, and Perplexity (HubSpot).
For category context, read what AI visibility means.
Semrush states the formula as:
AI share of voice = (your AI mentions ÷ total AI mentions across all brands in your category) × 100 (Semrush).
The numerator counts mentions of your brand. The denominator counts mentions across all brands in the category. The result is a competitive ratio, not a raw mention count (Semrush).
The Semrush example can be read in three parts:
The category total is essential to the result. Ten mentions do not produce a share of voice percentage on their own. The calculation needs the total mentions across the brands included in that category (Semrush). HubSpot's 25-of-100 example makes the same point: the percentage comes from a brand count relative to the full mention count (HubSpot).
The formula becomes inspectable when the prompt set and mention log sit beside the ratio. That is the purpose of the Jungle Roots proof wall: show the observation behind the claim instead of presenting a percentage without its evidence.
These terms belong in separate columns. Semrush distinguishes all four in its measurement guide (Semrush).
| Metric | What to record in the ledger | What the record establishes | What it does not establish |
|---|---|---|---|
| Mention | The brand named in the answer. | The brand appeared in that observed answer. | That the brand's website was cited. |
| Citation | The page visibly attributed or linked in the answer. | That page was cited in that observed answer. | That the cited domain was also the brand named in the answer. |
| Source | A webpage or document an AI system retrieves to build an answer (Semrush). | The AI system retrieved it to build an answer. | A visible citation or brand mention. |
| AI share of voice | Your AI mentions, total mentions across all brands in the category, and the resulting percentage (Semrush). | Your share of the observed category mentions. | A citation rate or traditional share of voice. |
A mention does not establish a citation. A citation does not establish that the cited domain is the named brand. Keeping the rows separate prevents one observation from being reported as another.
The ledger should therefore answer different questions with different fields:
This distinction also separates AI share of voice from traditional share of voice. HubSpot says traditional share of voice covers channels such as advertising, organic search, and social media, while AI share of voice focuses on answer engines (HubSpot). For the search and answer-engine split, see SEO vs GEO.
Use a fixed prompt set and preserve the ledger behind the ratio. On our own portfolio, the operator method is:
This protocol gives the percentage a visible boundary: the prompt set in the ledger. It also lets another reader inspect which observations entered the numerator and denominator. The broader AI search engine optimization strategies for consulting firms guide applies the same evidence boundary to page publishing and citation measurement.
The order makes the result auditable. Writing the prompt set first defines which responses belong in the observation. Logging each response preserves the mentions used in the calculation. Counting both the brand numerator and category denominator exposes the inputs to the ratio. Separate citation and source fields stop those records from being folded into the mention count.
The final statement should retain that boundary. Instead of presenting the percentage as a universal property of the brand, state that it is the brand's share of mentions in the defined category and prompt set. The prompt set, mention log, numerator, denominator, citation fields, and source fields then show exactly what was counted.
Semrush says its Brand Performance report calculates AI share of voice using mention count and how high the brand appears (Semrush). That is a Semrush product method. The operator protocol here does not imply that Jungle Roots uses Semrush. It applies Semrush's published mention formula to an inspectable prompt set and ledger.
For a consulting-firm application of this measurement, see AI visibility for consulting firms. If the ledger shows no mention for the firm, why ChatGPT never mentions your firm frames the visibility problem.
The supported statement is: your brand held a stated share of the mentions observed in the defined category and prompt set. Other conclusions require their own evidence.
AI share of voice also does not prove why an answer engine named a brand, why a page was cited, or whether a named brand's own site supplied the source. The approved definitions describe observed visibility, mentions, citations, and sources. They do not turn one record into proof of another.
It does not prove performance in advertising, organic search, or social media either. HubSpot places those channels under traditional share of voice and distinguishes them from answer-engine share of voice (HubSpot). Report the answer-engine ratio for the answer-engine observation, and keep claims about other channels outside that result.
Choose the comparison set before collecting results, then apply one observation frame to every brand. Semrush defines AI share of voice through AI mentions relative to all brands in the category. HubSpot defines it through how frequently and prominently a brand appears relative to competitors. Those definitions make the comparison set part of the measure. They do not, by themselves, define the operating protocol below.
Jungle Roots methodology: a fair comparison uses the same prompts, engines, capture date, and inclusion rule for every brand. The prompt set determines which questions enter the sample. The engine set determines which answer systems enter it. The capture date fixes when responses were observed. The inclusion rule decides what counts as a brand mention. If any of these differ by brand, the resulting shares do not describe the same observation frame.
| Method element | What to declare | Why it must stay consistent |
|---|---|---|
| Prompts | The exact prompt set | A changed set creates a different sample |
| Engines | Every engine included | Each brand must face the same engine set |
| Date | The capture date | Observations must share the same time frame |
| Inclusion rule | What counts as a mention | The numerator must follow one rule |
| Comparison set | Every brand in the denominator | Adding or removing a brand changes the share |
Use a frozen method when you want to compare periods. Keep the prompt set, engines, inclusion rule, and competitor set unchanged, and disclose the capture date for each period. This preserves a stable denominator and makes the comparison easier to interpret.
If the market or reporting need requires a method change, restate the earlier period under the new method when possible. Label the restatement clearly. If you cannot restate it, report the new result as a new series rather than presenting it as directly comparable. A practical AI visibility reporting template can keep the frame, counts, and method notes together. The /proof page shows where supporting evidence can be published.
Before reporting a result:
This protocol keeps the percentage tied to what was actually observed. Zero mentions remains a result for the declared sample, not proof that an engine never mentions the brand. Likewise, a rise or fall shows a change within the stated frame. It does not establish why the change happened.
AI share of voice percentages have meaning only within a declared sample. The published mention formula divides one brand's AI mentions by total AI mentions across all brands in the category, then multiplies the result by 100 (Semrush). That denominator defines the observation frame. Change the brands or category in the frame, and you are no longer interpreting the same comparison.
| Score | Meaning inside the declared sample | What it does not mean |
|---|---|---|
| 0% | Zero of the counted category mentions belonged to the brand in that observation frame (Semrush). | The brand is never mentioned. |
| 50% | Half of the counted category mentions belonged to the brand in that observation frame (Semrush). | The brand owns half of all answers, buyers, citations, or market demand. |
| 100% | Every counted category mention in that observation frame belonged to the brand (Semrush). | The brand has universal or permanent dominance. |
None of these percentages is universally good. Semrush and HubSpot define AI share of voice relative to the brands included in a category or competitor comparison, but they do not publish a universal good percentage (Semrush; HubSpot). A useful result is one measured against a frozen, relevant comparison frame and tied to a stated business question. This is Jungle Roots methodology, not a vendor benchmark.
For example, a 50% score can answer a narrow question about a declared category sample. It cannot answer a broader question about buyer preference or market demand because those measures are outside the mention formula (Semrush). The same limit applies at 0% and 100%. Read each score as arithmetic about the counted mentions, not as a claim about every answer engine or future observation.
Use this checklist before interpreting a score:
No. AI share of voice is a competitive ratio based on brand mentions across a category (Semrush). AI visibility asks whether and how a brand appears in answer-engine responses (Jungle Roots).
No. Semrush distinguishes an AI mention from a citation and a source (Semrush). Record the named brand, cited page, and exposed source in separate fields.
Yes. The Semrush formula uses brand mentions in its numerator, while Semrush treats a source as a separate term (Semrush). A source appearance is therefore not automatically a brand mention.
HubSpot describes AI share of voice across answer engines such as ChatGPT, Gemini, and Perplexity (HubSpot).
Publish the prompt set, mention log, numerator, denominator, and separate citation and source fields. The proof wall is where Jungle Roots presents dated evidence and measurement definitions.
Need this ledger applied to your firm? Get the operations audit.
