Written by: Mariana Fonseca, Editorial Team, AI Growth Agent
Key Takeaways
- AI Share of Voice (AI SOV) tracks your brand’s percentage of mentions in AI-generated answers versus competitors within a fixed prompt set and time window.
- Multiple formula variants exist, including mention-based, position-weighted, and citation-based. Each produces different scores from the same dataset, so choosing one and sticking with it matters for reporting.
- A complete AI visibility dashboard pairs SOV with four supporting metrics: Mention Rate, Citation Rate, Prompt Coverage, and Competitor Gap. Together they show real competitive position.
- Position-weighted SOV using harmonic decay (1/position) reflects real prominence better than raw mention counts, because first-position mentions carry far more value.
- AI Growth Agent maps your full prompt universe and isolates the metrics that matter for your category. Schedule a consultation session to benchmark your current performance.
The Five-Metric AI Visibility Dashboard
AI SOV on its own gives an incomplete picture. A reliable measurement system tracks four supporting metrics alongside the core SOV formula.
Mention Rate is the percentage of all sampled prompts where your brand appears by name, regardless of competitors. MaxAEO distinguishes it from SOV directly. A brand can hold a 30% mention rate and only an 18% share of voice. The calculation is: (answers naming your brand ÷ all answers sampled) × 100. Cintra does not publish a specific numeric benchmark for a good AI mention rate.
Citation Rate measures how often your domain is explicitly linked as a source in AI responses. TurboAudit notes it correlates more strongly with revenue than simple mention counts. The calculation is: (prompts citing your URL ÷ total prompts) × 100.
Prompt Coverage tracks the percentage of your defined prompt set that triggers any mention of your brand. Texta flags coverage below 40% as a sign of major gaps in AI visibility. Coverage of 80% or higher means the brand appears consistently in relevant AI responses.
Competitor Gap (also called Inverse SOV or Missed Prompts) is the percentage of relevant prompts where your brand does not appear. TurboAudit finds it more actionable than standard SOV because it surfaces specific content gaps. The following table summarizes how each metric is calculated and whether position weighting applies.
| Metric | Formula | Raw vs. Weighted |
|---|---|---|
| Mention Rate | (Brand answers ÷ All answers) × 100 | Raw only |
| Citation Rate | (Cited prompts ÷ Total prompts) × 100 | Raw only |
| AI SOV | (Brand mentions ÷ All tracked mentions) × 100 | Both, weighted applies 1/position decay |
| Prompt Coverage | (Prompts with brand ÷ Total prompt set) × 100 | Raw only |
| Competitor Gap | (Prompts without brand ÷ Total prompt set) × 100 | Raw only |
How Position-Weighted SOV Works in Practice
This worked example uses three brands across 10 prompts on a single engine and applies harmonic decay (1/position) as documented by SEO Hacker.
The position weights used are:
- Position 1: weight = 1.00
- Position 2: weight = 0.50
- Position 3: weight = 0.33
- Position 4+: weight = 0.25 or lower
The step-by-step calculation follows a simple sequence.
- Run your prompt set, with a minimum of 30 prompts per Maximus Labs methodology, across your chosen engines.
- Record every brand mentioned in each answer and its position in the response.
- Assign the position weight to each mention using 1/position.
- Sum each brand’s weighted mentions across all prompts.
- Divide each brand’s weighted total by the pool’s combined weighted total and multiply by 100.
The following table shows a worked numerical example across 10 prompts.
| Brand | Raw Mentions | Weighted Mentions (1/position) | Position-Weighted SOV |
|---|---|---|---|
| Brand A (avg. position 1) | 8 | 8.00 | ~52% |
| Brand B (avg. position 2) | 7 | 3.50 | ~23% |
| Brand C (avg. position 3) | 6 | 1.98 | ~13% |
Brand A’s raw mention count leads by only one mention over Brand B, yet its position-weighted SOV is more than double. Weighted measurement highlights who AI actually recommends first, while raw counts blur that reality.
MaxAEO references an alternative weighting approach that is less mathematically strict but easier to explain to executives. Both approaches are valid. The key is to keep the chosen method consistent across reporting periods.
Interpreting Scores with the 30% Rule and Benchmarks
Parity provides the baseline reference point for any AI SOV score. Parity equals 100 divided by the number of brands tracked. In a five-brand competitive set, parity is 20%. MaxAEO defines an excellent result as 2x parity or more and formalizes this as the parity index: your SOV divided by (100 ÷ number of brands).
The 30% rule marks the threshold where a brand’s AI SOV begins to signal market-leader candidacy. Rankeo defines the interpretation bands as:
- Below 5%: niche or emerging brand
- 5% to 15%: established tier-2 brand
- 15% to 30%: market leader candidate
- 30% and above: dominant
Category-specific benchmarks vary widely. Foglift’s analysis of thousands of scans across industries produces these ranges.
- SaaS and B2B Software: leader SOV 40% to 55%, mid-tier 15% to 25%, laggard below 8%
- E-Commerce and DTC: leader SOV 30% to 45%, mid-tier 12% to 20%, laggard below 6%
- Financial Services: leader SOV 35% to 50%, mid-tier 15% to 22%, laggard below 7%
Cassie Clark Marketing adds maturity-stage context. In early-stage or emerging categories, 30% to 50% already counts as strong. In competitive or mature categories, 40% to 60% is defensible. For branded prompts such as “Brand X pricing,” 50% to 80% is expected. For non-branded prompts, 30% to 60% is often a strong result.
MaxAEO’s benchmark data from 2.4 million AI answers sampled in 2026 across 412 competitive sets in 38 B2B software categories shows that median SOV values differ by category. Category context is essential before interpreting any single score.
Finding Missed Prompts with Coverage and Competitor Gap
Most monitoring tools track a capped set of prompts selected in advance. That architecture measures only the slice of the market a brand already thought to ask about. Real customer queries for finding, comparing, and evaluating brands in AI search extend far beyond any pre-selected list.
Teams that want the full universe need a different approach. Foglift does not specify a minimum prompt count for statistically meaningful AI SOV results in its guides. Rankeo sets a standard measurement setup at 100 prompts across five engines that reflect real customer questions in the vertical.
Competitor gap analysis compares your brand’s performance across mention rate, position, sentiment, and citation coverage against named competitors. The goal is to reveal white space in the AI consideration set.
- Define a prompt set that spans all buyer journey stages, including awareness, comparison, and decision. This approach measures the full journey, not just bottom-funnel queries.
- Run each prompt across ChatGPT, Perplexity, Gemini, Claude, and Grok, as Rankeo identifies these five as the standard engines for benchmark-grade measurement.
- Record every brand mentioned per prompt and its position. This data shows which brands AI considers relevant for each query.
- Identify prompts where competitors appear and your brand does not. These missed prompts represent the clearest opportunities for visibility gains.
- Cluster missed prompts by topic to identify content gaps that, once addressed, close the competitor gap.
Topify notes that high-performing brands maintain citation coverage across industry publications, YouTube transcripts, Wikipedia, and review sites, not just their own domain. A competitor gap analysis that surfaces which third-party sources drive competitor citations is more actionable than one that only counts brand mentions.
AI Growth Agent maps the entire universe from real-time Google and ChatGPT data, running more than 3,000 searches every week to refresh the snapshot. Prompt count is never a billed metric, so the competitive gap analysis reflects the actual market instead of a capped approximation.
Why Weekly Tracking and Multi-Engine Mapping Matter
AI citations shift constantly. ASEO and Profound 2025 to 2026 data shows citations vary 40% to 60% month over month for the same query. Fishkin and O’Donnell’s approximately 3,000-run study in 2026 found less than 1-in-1,000 odds of two AI runs producing the same ordered brand list. Single-shot or quarterly manual checks miss most citation drift between checkpoints.
The recommended tracking cadence by measurement tier is:
- Weekly: leading indicators including AI Visibility Rate by platform, new citation wins, and AI bot crawl coverage, as defined by Machine Relations
- Bi-weekly: core metrics including Share of Citation, Citation Rate trend, and competitive displacement rate
- Monthly: full audits of 30 to 50 prompts across all engines, per Shadow’s 2026 framework
Multi-engine tracking is essential. Profound’s 2026 analysis of 680 million citations found only about 11% of domains are cited by both ChatGPT and Perplexity. Each engine functions as a separate measurement game. A brand achieving 67% SOV on Claude and Perplexity but 0% on OpenAI models, as documented in a Finnish ecommerce study by Niko Alho, would remain invisible in a single-engine report.
Incremental visibility isolation completes the picture. Reporting that blends new content performance with existing brand visibility cannot prove whether your investment works. AI Growth Agent publishes into a separate environment and reports week over week what it generated. The system cross-references bot traffic, Google Search Console, and citation data to produce a defensible number instead of a blended estimate.
Why Raw SOV Needs Sentiment and Accuracy Checks
Raw AI SOV has structural limitations that teams must address before presenting numbers to executive stakeholders.
The first limitation is sentiment blindness. Lumentir documents a scenario where a brand holds 35% share of voice built entirely on negative mentions, complaints, criticism, or regulatory concerns. Raw counts miss that nuance. Shadow’s 2026 framework identifies sentiment classification as critical because a high mention rate with negative sentiment creates toxic visibility that harms the brand.
The second limitation is the hidden denominator problem. Dan Taylor, Head of Technical SEO at SALT.agency, describes this as extrapolating from small, fixed prompt sets to an effectively infinite universe of possible queries. That pattern produces structurally unreliable assessments. Nick Lafferty’s June 2026 analysis of 48,589 citations found that 95% of product titles appeared in under 30% of runs of the same prompt.
The third limitation is formula divergence. The same brand and dataset can produce 20% mention-based SOV, 16.8% position-weighted SOV, and 31.4% citation-based SOV. Digital Applied confirms there is no industry-standard formula as of mid-2026.
The paired metrics that provide fuller context are:
- Sentiment score, which tracks positive, neutral, or negative tone of AI-generated mentions and appears in tools including Peec, AthenaHQ, and HubSpot
- Brand accuracy, which measures whether AI models describe your brand correctly in terms of positioning and capabilities, identified by Yes Optimist as a Tier 1 visibility metric
- Citation absorption, which distinguishes content that drives the narrative from content cited only as a footnote. Yao et al.’s 2026 study of 602 prompts finds citation and absorption function as discrete stages.
Automated systems are required to track these paired metrics at scale. Manual tracking of sentiment across hundreds of prompts and multiple engines introduces human judgment variability that blocks reliable trend detection, as Lumentir documents. Onely notes that automated monitoring reduces cost at enterprise scale while improving coverage of competitive changes.
Turning AI SOV Metrics into Automated Growth
Measurement without action remains a dashboard exercise. Many teams know their AI SOV yet struggle to change it. Monitoring tools report the number. They do not create content, publish it with full technical SEO, or keep it updated as the market shifts.
The conversion from metrics to action requires a system that follows a clear sequence. First, it must map the full prompt universe, not a capped subset, using real-time AI Overview and ChatGPT data to decide which queries deserve attention. Once the system knows which prompts matter, it must produce authoritative content for each identified gap, with every claim validated against primary sources instead of model training data.
Next, that content needs to be published with the full technical and agentic SEO stack live on day one. The stack includes schema, Blog MCP, llms.txt, llms-full.txt, agent discovery via /.well-known/, and bot tracking that shows exactly when ChatGPT cites the content. Finally, the system must self-heal content over time so the brand’s presence does not decay as model updates and index refreshes change citation patterns.
Capped monitoring tools stop at the first step and only tell you the score. AI Growth Agent runs the full loop and changes the score.
The impact of this architecture is measurable. Across the first twelve weeks, AI Growth Agent clients average more than 12,000 additional AI citations and mentions, over 100,000 additional bot visits, and a lift of more than 20% in impressions. Breadless grew from 387,000 to 12.3 million Google Search Console impressions in six months, a 30x lift, with ChatGPT citing eatbreadless.com over 45,000 times per month. Leva Sleep closed $40,000 to $50,000 in deals in under three weeks from buyers who discovered the brand through AI Growth Agent content.
The engine runs headless. It operates without adding headcount, replaces the SEO agency, the content tool, the GEO monitor, the schema plugin, and the analytics stack, and reports the incremental visibility it generates week over week so the dashboard reflects what the engine actually produced.
Frequently Asked Questions
How do you measure AI share of voice?
Teams measure AI share of voice by running a fixed prompt set across a defined list of AI engines, recording every brand mentioned in each answer, and dividing the brand’s total mentions by the combined mentions of all tracked brands, then multiplying by 100. The base formula is: AI SOV (%) = (your brand’s mentions ÷ total mentions of all tracked brands) × 100. For more accurate competitive assessment, many teams apply position weighting using harmonic decay (1/position), so a first-position mention counts as 1.00, second position as 0.50, third as 0.33, and so on. A practical setup runs each prompt a minimum of five times per engine to account for non-deterministic model outputs and tracks at least ChatGPT, Perplexity, Gemini, Claude, and Grok to avoid single-engine blind spots. Teams then pair the core SOV score with mention rate, citation rate, prompt coverage, and competitor gap analysis. A 100 to 200 prompt panel run weekly functions as a minimum viable tracking system in active categories.
What is the 30% rule in AI?
The 30% rule in AI share of voice marks the threshold where a brand’s SOV score begins to signal market-leader candidacy in a competitive category. Brands below 15% usually sit in an established tier-2 position. Brands between 15% and 30% operate as market leader candidates. Brands above 30% are considered dominant within their tracked competitive set. The rule functions as a benchmark rather than a guarantee, because category maturity, the number of brands tracked, and the prompt set used all affect what a given percentage means. In early-stage or emerging categories, 30% to 50% already counts as strong. In competitive or mature categories, 40% to 60% is defensible. The 30% threshold works best as a directional target for brands moving from tier-2 to leadership and should always be interpreted against the parity baseline of 100 divided by the number of brands in the tracked set.
What is a good share of voice percentage?
A good AI share of voice percentage depends on category, competitive set size, and brand maturity. The universal baseline is parity, defined as 100 divided by the number of brands tracked. Anything above parity is positive, and 2x parity or more is excellent. Category-specific ranges appear earlier in this article, but as a quick reference, SaaS leaders typically hold 40% to 55%, e-commerce leaders 30% to 45%, and financial services leaders 35% to 50%. For branded prompts such as “Brand X pricing,” 50% to 80% is expected. For non-branded category prompts, 30% to 60% is often a strong result. In categories with five to eight serious competitors, 20% to 30% signals strong visibility and 35% to 50% indicates market-leader territory. Scores below 10% in a five-plus competitor category mean the brand rarely enters AI-generated buyer shortlists.
What does 50% share of voice mean?
A 50% AI share of voice means your brand accounts for half of all tracked brand mentions across the defined prompt set and competitive set. In practical terms, AI engines name your brand in roughly one out of every two answers that mention any tracked competitor. At 50%, a brand acts as a primary reference in AI answers and meaningfully shapes the category conversation. Context still matters. In a two-brand competitive set, 50% equals parity. In a ten-brand competitive set, 50% is dominant. In mature, highly competitive categories, 50% often indicates consolidation around a small group of trusted sources. In emerging categories, 50% can be achievable within the first few months of a systematic content and visibility program. The score should always be read alongside sentiment and accuracy data, because 50% share of voice built on negative or inaccurate mentions creates a misleading picture of brand health.
