Measuring Brand Share in AI Answers
Short answer
To calculate brand share in generative AI responses, divide the count of your brand mentions by total category brand mentions across a standardized prompt set, multiplying by 100 ([Arxiv](https://arxiv.org/pdf/2606.20065v1); [Agilebrandguide](https://agilebrandguide.com/wiki/agentic-commerce/share-of-model-som)). Across more than 100,000 prompt responses analyzed on Ranqo, global brands achieved an initial baseline appearance rate of 73% ([Arxiv](https://arxiv.org/pdf/2606.20065v1)), mid-market brands captured 44% ([Arxiv](https://arxiv.org/pdf/2606.20065v1)), and niche brands secured 11% ([Arxiv](https://arxiv.org/pdf/2606.20065v1)). Tracking requires auditing both pure text mentions and source citations across multi-model query evaluations ([Arxiv](https://arxiv.org/pdf/2606.20065v1)).
“While SEO focuses on own-site content, in many cases, a brand's own sites only comprise 5 to 10 percent of the sources that AI-search references”
— McKinsey & Company
“A single blended visibility percentage can hide the fact that a product is visible for broad category prompts but absent from the framework-specific prompts that actually drive adoption.”
— Ben Williams, CEO, DevTune
Traditional organic search metrics cannot reflect how generative models synthesize recommendations. When generative systems construct conversational answers, brand presence shifts from search rankings to explicit text recommendations and linked source citations. Measuring brand presence in this environment requires calculating mention frequency and model share across structured query sets rather than tracking traditional index positions.
Formulas and Methods for Calculating Brand Visibility and Share of Model#
Brand visibility reflects how frequently an AI assistant outputs a company name across a defined evaluation set. Ranqo measures single-engine brand visibility as the mention rate, dividing the prompts where a brand is named by the total prompts tested and multiplying by 100 (Arxiv):
Engine Visibility Rate (%) = (Brand Mentions / Total Prompts) × 100
Measuring performance across diverse systems requires a weighted average. In Ranqo's tracking, cross-engine visibility is calculated through assigned platform weights: ChatGPT at 0.30, Gemini at 0.20, Perplexity at 0.20, Claude at 0.20, and Grok at 0.10, renormalizing across active platforms so the weights sum to 1.0 (Arxiv).
Share of Model (SoM) defines the percentage of AI assistant answers or recommendations within a product category that mention or recommend a specific brand (CDP). Industry frameworks evaluate SoM through live responses, calculating the proportion of brand mentions captured across category prompts (Agilebrandguide). The primary calculation evaluates brand mentions against total category brand mentions (Agilebrandguide):
Share of Model (%) = (Brand Mentions in Category Responses / Total Brand Mentions in Category Responses) × 100
Alternatively, competitive share of voice isolates verified market rivals. Ranqo computes this by dividing brand mentions by the sum of brand mentions and recurring competitor mentions, excluding one-off hallucinated names (Arxiv).
| Metric | Primary Formula | Operational Function |
|---|---|---|
| Single-Engine Visibility | (Brand Mentions / Total Prompts) × 100 (Arxiv) |
Evaluates raw mention frequency on one engine. |
| Cross-Engine Visibility | Platform-weighted average across active engines (Arxiv) | Normalizes presence across multi-model search stacks. |
| Share of Model (SoM) | (Brand Mentions / Total Category Mentions) × 100 (Agilebrandguide) |
Assesses brand share relative to category competitors. |
| Competitive Share of Voice | Brand Mentions / (Brand Mentions + Recurring Competitor Mentions) (Arxiv) |
Removes hallucinated entities to assess verified rivals. |
Sentiment score provides an additional qualitative weight. Ranqo scores sentiment using (positive + 0.5 neutral) / mentions × 100, where all-positive output scores 100, neutral scores 50, and negative scores 0 (Arxiv). Under this weighting, a negative mention penalizes a brand more heavily than a neutral mention (Arxiv).
Structuring and Versioning Prompt Evaluation Panels#
Monitoring generative retrieval requires structured prompt taxonomies that mirror real buyer research journeys. Ranqo segments queries into six distinct prompt categories: discovery, problem/solution, use case, comparison, expert, and brand research (Arxiv).
Alternative evaluation structures organize prompts by corporate attributes. The 5W AI Visibility Index evaluates brands across four distinct operational categories: Category A for business description (15 prompts), Category B for founder and leadership (12 prompts), Category C for competitive positioning (18 prompts), and Category D for buyer intent (15 prompts) (Everything PR).
Longitudinal tracking demands strict query control to prevent measurement artifacts. Analysts must keep the core prompt set fixed across tracking cycles, rotating no more than 10% to 15% (Business) of the evaluation panel per quarter (Business). This controlled rotation preserves usable historical trend lines without introducing wording drift (Business).
- Define Category Taxonomy
- Fix Core Prompt Panel
- Execute Daily API Audits
- Rotate 10-15% ([Business](https://business.daily.dev/resources/measure-brand-share-of-voice-ai-answers)) Quarterly
The SEO and AI Search Marketing discipline exists to turn conversational queries into measurable, repeatable retrieval workflows rather than anecdotal search experiments.
Citations by Web Source Type and Page Format#
Generative retrieval pipelines gather information from third-party ecosystems rather than primary corporate channels alone. Large-scale tracking across more than 100 brands on Ranqo revealed that corporate websites represent about 78% (Arxiv) of citations generated by AI search engines (Arxiv). However, a brand's own websites comprise only 5 to 10 percent of the sources referenced by AI search engines (Mckinsey).
Outside of corporate websites, models prioritize specific non-corporate platforms. According to an analysis published by Ranqo, among non-corporate sources, YouTube is cited most frequently, followed by Reddit, editorial media, and Wikipedia (Arxiv).
Specific editorial page formats deliver outsized citation frequency. At the URL level, ranked "best-of" listicles represent the most-cited content format, making up about 21% of all citations (Arxiv). Because a single comparative roundup can surface a product across numerous generative responses, third-party listicles represent high-leverage assets in generative search (Arxiv).
Baseline Visibility Rates Across Brand Stature Tiers#
Initial visibility in generative responses scales directly with established market stature. Between March and May 2026, an analysis of more than 100,000 prompt responses across more than 100 brands on Ranqo demonstrated that initial baseline appearance rates follow a clear three-tier ladder (Arxiv):
- Global household names (such as Stripe and Nike) appeared in 73% (Arxiv) of relevant AI answers on their initial tracking run (Arxiv).
- Established mid-market and regional brands (such as Olipop and Klaviyo) appeared in 44% (Arxiv) of relevant responses (Arxiv).
- Niche and small brands appeared in 11% (Arxiv) of relevant answers (Arxiv).
Each stature tier falls approximately 30 percentage points below the prior group (Arxiv). For mid-market and emerging brands, overcoming this structural distribution gap requires actively earning presence across external reference sources rather than relying purely on primary domain authority.
Observed Variance in Citations and Sentiment Framing Over Time#
Citation indexes display substantial instability across repeated measurement runs. A tracking study of 2,500 prompts across Google AI Mode and ChatGPT conducted with the Semrush AI Visibility Index found that between 40% and 60% (Searchengineland) of cited sources change from month to month (Searchengineland). This continuous source rotation indicates that static ranking assumptions fail inside dynamic retrieval engines.
Sentiment framing exhibits even greater volatility than source citation lists. Research on the Ranqo platform found that whether an AI model frames a brand positively or negatively flips approximately 6.7 times (Arxiv) more frequently than whether the brand is mentioned at all (Arxiv).
Because narrative framing fluctuates rapidly while underlying entity recognition remains stable, marketing leaders should track core mention presence and external source citations independently of model tone.
Next Steps for AI Visibility Audits#
Evaluating brand share across generative platforms requires moving from isolated prompt tests to structured, verifiable measurement pipelines. To assess your current generative presence, execute the following audit steps:
- Assemble a fixed panel of 50 to 100 category prompts divided across discovery, problem/solution, and competitive comparison categories (Arxiv).
- Query target AI platforms through official API endpoints, clearing chat sessions between iterations to eliminate conversational history bias.
- Calculate your baseline Engine Visibility Rate and Share of Model, separating recurring industry competitors from non-recurring entities (Arxiv; Agilebrandguide).
- Audit external citations to verify if your product appears within the top third-party corporate domains and comparative listicles - which generate roughly a fifth of citations, as detailed above (Arxiv).
- Re-run prompt panels across consistent monthly cadences, keeping prompt panel churn below 15% quarterly to protect measurement continuity as outlined above (Business).
For organizations seeking rigorous attribution, Porter Labs applies The Woof Back methodology: finding the core questions a market asks, answering them with sourced content on the client's own domain, checking daily whether ChatGPT, Perplexity, Gemini, and Claude cite the client, and tracing each inquiry back to the article that produced it through a first-party pixel and CRM records. To evaluate your baseline standing across leading answer engines, schedule a structured AI search visibility review.
| AI Engine | Model Class | Web-Search Policy | Citation Surface | Ranqo Cross-Engine Weight |
|---|---|---|---|---|
| ChatGPT | GPT-5 | Weekly cooldown per brand | annotations.url_citation | 0.30 |
| Gemini | Gemini-3 | Always grounded | groundingChunks[].web | 0.20 |
| Perplexity | Sonar | Always grounded (search-native) | search_results array | 0.20 |
| Claude | Claude Sonnet | Weekly cooldown per brand | text-block citations[] | 0.20 |
| Grok | Grok-3 | Always-on retrieval | citations[] + inline markdown | 0.10 |
| Measurement Dimension | Observed Metric / Range | Key Characteristics | Source |
|---|---|---|---|
| Global Household Brands | 73% visibility | Baseline appearance rate on initial tracking runs (e.g., Stripe, Nike) | Ranqo / Arxiv |
| Mid-Market & Regional Brands | 44% visibility | Baseline appearance rate on initial tracking runs (e.g., Olipop, Klaviyo) | Ranqo / Arxiv |
| Niche & Small Brands | 11% visibility | Baseline appearance rate on initial tracking runs (~30 points below mid-market) | Ranqo / Arxiv |
| Corporate Web Citations | ~78% of citations | Dominant source pool across all citations from AI engines | Ranqo / Arxiv |
| Ranked Best-Of Listicles | ~21% of citations | Highest-leverage individual page format surfacing products in answers | Ranqo / Arxiv |
| Cited Source Volatility | 40% to 60% change | Monthly rate of citation turnover across Google AI Mode and ChatGPT | Semrush / Searchengineland |
AI Brand Visibility & Share of Model Calculator
Calculates brand visibility rate across an AI evaluation panel and Share of Model (SoM) against total category brand mentions using standard generative search audit formulas.
Brand Visibility Rate:
Share of Model (SoM):
Prompts Without Brand Mention:
Frequently Asked Questions
Which metrics best quantify recommendation sentiment in LLM outputs?
Ranqo calculates sentiment score using the formula (positive + 0.5 neutral) / mentions × 100. This assigns 100 to all-positive responses, 50 to neutral responses, and 0 to negative responses, ensuring a negative mention carries twice the penalty of a neutral mention (Arxiv).
Sources
- arxiv.org (via Perplexity sonar-pro) - Arxiv (2026-10-06)
- AI Share of Voice - Definitions, FAQs & How HubSpot Helps - Hubspot (2026-10-06)
- New front door to the internet: Winning in the age of AI search - Mckinsey (2026-10-06)
- Generative engine optimization (GEO): How to win AI mentions - Searchengineland (2026-10-06)
- Share of Model (SoM) - The Agile Brand Guide® - Agilebrandguide (2026-10-06)
- AI Visibility Index: Methodology & Research Framework - Everything PR (2026-06-09)
- What Is Share of Model? The AI Visibility Metric | CDP.com - CDP (2026-10-06)
- How to measure your brand's share of voice in AI answers | daily.dev Ads - Business (2026-10-06)