By maxaeo.ai ｜ Published 2026-11-12 ｜ Updated 2026-11-12
The AI share of voice calculation formula measures how much of the competitive conversation your brand owns inside AI-generated answers. The most defensible baseline is:
AI Share of Voice (SOV) = Your brand mentions ÷ Total tracked brand mentions × 100
The formula is simple. The difficult part is defining “mentions,” choosing a consistent prompt set, handling recommendations and citations, and preventing one unusually long AI answer from distorting the result.
What is AI share of voice?
AI share of voice is the percentage of tracked brand presence attributed to your company compared with competitors across a defined set of AI answers.
For example, suppose a SaaS team runs 100 buyer prompts across ChatGPT, Perplexity, Gemini, and Claude. The answers contain:
| Brand | Mentions |
|---|---|
| Brand A | 40 |
| Brand B | 35 |
| Brand C | 15 |
| Brand D | 10 |
| Total | 100 |
Brand A’s AI SOV is:
40 ÷ (40 + 35 + 15 + 10) × 100 = 40%
This is a competitive, zero-sum metric. When one brand gains share, at least one other tracked brand loses relative share.
That makes SOV different from visibility or mention rate. A brand can appear in 40% of answers while receiving only 20% of all brand mentions if many answers list several companies.
The three formulas teams often confuse
Search results commonly use “share of voice” to describe several related metrics. They should not be treated as interchangeable.
1. Mention rate
Mention Rate = Answers containing your brand ÷ Total answers × 100
If your brand appears in 30 of 100 answers, its mention rate is 30%.
Mention rate is useful for measuring coverage, but it is not necessarily competitive SOV. Multiple competitors can appear in the same answer, so their mention rates may add up to more than 100%.
2. Competitive AI share of voice
Competitive SOV = Your brand mentions ÷ All tracked brand mentions × 100
This is the cleanest formula when the question is: “How much of the category conversation belongs to us?”
It requires a fixed competitor set and a consistent counting rule.
3. Citation share
Citation Share = Your cited URLs or domains ÷ All tracked cited URLs or domains × 100
Citation share measures source authority rather than brand naming. A brand may be mentioned without its website being cited. Conversely, a research report or comparison page may be cited without the brand receiving a prominent recommendation.
A reliable AI visibility report should show mention rate, competitive SOV, recommendation rate, and citation share separately.
How to calculate AI SOV step by step
Step 1: Define the measurement scope
Write down the variables before collecting answers:
- AI engines included
- Country and language
- Prompt categories
- Competitor list
- Measurement period
- Counting rules
- Whether citations and recommendations receive extra weight
Do not combine ChatGPT answers from one date with Perplexity answers from another period unless the reporting window is clearly documented.
Step 2: Build a buyer-focused prompt set
A useful prompt set should represent different stages of product research:
- Category prompts: “What are the best project management tools?”
- Use-case prompts: “What is the best project management software for remote SaaS teams?”
- Comparison prompts: “Product A vs Product B for a growing startup”
- Problem prompts: “How can a SaaS team reduce customer churn?”
- Trust prompts: “Which tools have strong customer reviews and reliable support?”
Use the same prompts for every tracked competitor. Changing the prompt set during a reporting period makes the trend difficult to interpret.
Teams can also convert existing SEO keywords into AI-style buyer prompts. This connects traditional search demand with answer-engine measurement instead of creating a disconnected reporting system.
Step 3: Count brand presence consistently
For each answer, record:
- Whether the brand was mentioned
- Number of brand mentions
- Whether it was recommended
- Position in a list or answer
- Whether the brand was described positively, neutrally, or negatively
- Whether a source was cited
- Which domain or page was cited
A practical raw dataset might use one row per answer:
| Date | Engine | Prompt | Brand | Mentioned | Recommended | Position | Cited |
|---|---|---|---|---|---|---|---|
| 2026-11-12 | ChatGPT | Best CRM for SaaS | Brand A | 1 | 1 | 2 | 1 |
This structure prevents a common error: treating a brand that appears once in a long list as equal to a brand that is presented as the top recommendation.
Step 4: Apply the baseline formula
If your brand receives 68 mentions and all tracked brands receive 250 mentions:
68 ÷ 250 × 100 = 27.2% AI SOV
Report the denominator with the percentage. “27.2% SOV” is not reproducible unless readers know the number of prompts, engines, answers, and competitor mentions behind it.
A better framework: separate presence from influence
The baseline formula is useful, but it treats every mention equally. For SaaS buying decisions, that can hide important differences.
A stronger reporting model uses two layers:
Layer 1: Presence SOV
Presence SOV = Brand mentions ÷ Total brand mentions × 100
This answers: How often is the brand part of the category conversation?
Layer 2: Influence-adjusted score
Use a transparent weighted score for decision-oriented appearances:
Influence Score =
(1.0 × mentions)
+ (1.5 × recommendations)
+ (1.25 × top-three positions)
+ (1.25 × cited appearances)
Then:
Influence SOV =
Your Influence Score ÷ Total competitor Influence Scores × 100
The weights above are a practical starting framework, not a universal industry standard. The important principle is to publish the weights and keep them stable.
This approach adds information that a simple mention count misses. A brand with fewer mentions may still have stronger commercial visibility if it appears more often in top recommendations and trusted cited answers.
How to handle variability across AI engines
AI answers are not fixed search-result pages. Responses may change because of model updates, retrieval sources, location, language, prompt wording, or answer length.
For that reason, avoid interpreting a single response as a trend. Instead:
- Run the same prompt set repeatedly.
- Report results by engine before combining them.
- Keep language and geography consistent.
- Store the original answers for auditability.
- Use a rolling average for daily or weekly reporting.
- Flag prompts with unusually unstable results.
A useful combined score is an engine-normalized SOV:
Engine-Normalized SOV =
Average of each engine’s brand SOV
This prevents an engine with more prompts or more frequent sampling from dominating the total. If business importance differs by market, use an explicitly documented business weight rather than silently changing the denominator.
What should an AI SOV dashboard include?
A useful dashboard should answer both “how much?” and “why?” Include:
- Total AI mention rate
- Competitive share of voice
- Recommendation rate
- Average recommendation position
- Citation share
- Sentiment distribution
- Factual accuracy issues
- SOV by prompt category
- SOV by AI engine
- Competitor trend lines
- Cited domains and pages
The cited-source view is especially actionable. If competitors are repeatedly supported by review sites, comparison pages, technical documentation, or community discussions, the next optimization task is clearer than simply seeing a lower score.
For a broader measurement model, see this guide to AI search visibility dashboard metrics. For competitor-focused analysis, competitor AI mention tracking shows how to identify prompts where rivals appear and your brand does not.
Common questions about AI share of voice
Is AI share of voice the same as AI visibility?
No. AI visibility usually describes whether and how often a brand appears. AI SOV adds a competitive denominator, showing the brand’s relative share among tracked companies.
Can AI SOV be higher than 100%?
Competitive SOV should not exceed 100% when calculated from a shared denominator. A metric above 100% usually indicates that the tool is reporting a different measure, such as an index or weighted score.
Should recommendations count more than ordinary mentions?
Usually, yes—if the goal is to estimate buyer influence. However, weighted SOV should be reported separately from raw SOV, with the weighting method disclosed.
How many prompts are needed?
There is no universal number. A small diagnostic can reveal obvious gaps, while ongoing measurement should use enough prompts to represent major buyer intents, use cases, and comparison scenarios. Consistency matters more than an arbitrary large number.
How can SaaS teams monitor this without maintaining spreadsheets?
A platform such as MaxAEO’s AI visibility monitoring system can track brand mentions, competitive rankings, recommendation position, sentiment, and cited sources across eight AI engines with daily updates. A free diagnostic report can provide an initial view of the brand’s AI search presence.
Final takeaway
The most defensible AI share of voice calculation formula is:
Your brand mentions ÷ Total tracked competitor mentions × 100
Use mention rate to measure coverage, competitive SOV to measure relative presence, citation share to understand source authority, and influence-adjusted SOV to estimate recommendation strength. The metric becomes trustworthy when the prompt set, competitor universe, sampling period, and weighting rules are transparent and repeatable.