AI share of voice: comparing visibility across brands
AI share of voice describes a brand’s share of observed or modeled AI presence within a defined comparison group. Providers may use mentions, citations or weighted estimates. The result is meaningful only alongside the selected brands, data source and formula.
Also known as: AI share of voice · AI SOV
At a glance
- Share of voice is a comparative measure, not the same as mention rate.
- Adding a competitor can change your percentage without changing your own appearances.
- Modeled AI impressions are not actual human views or market share.
On this page

What is the share measured against?
Two boundaries define the comparison: the questions assessed and the brands included. Changing either can move the percentage. Share of voice is therefore not a permanent property of a brand.
A search competitor may not be a direct commercial competitor. Manufacturers, retailers and professional associations can appear in the same answer. Decide whether you are comparing similar providers, source visibility or every visible brand in a topic.
A simple calculation
Suppose a fictional study follows three brands and counts each brand at most once per answer. Your business appears in twenty answers, while the others appear in thirty and fifty.
Your mention-based share is 20 ÷ (20 + 30 + 50) = 20%. This is an illustrative method, not a universal definition used by every tool.
Add a fourth brand with fifty appearances and your share becomes 20 ÷ 150, or approximately 13.3%. Your twenty appearances have not declined. The denominator has expanded.
Several brands can occur in one answer. Their answer-level mention rates can consequently sum to more than 100%. That differs from dividing a shared total of mentions into shares that sum to 100%.
Why might a provider's chart differ?
Ahrefs' metric documentation describes AI Share of Voice using modeled impressions relative to tracked brands. Weighting and aggregation across platforms can produce a different figure from the simple count above.
Another report might use citations, domain occurrences or a proprietary visibility formula. When two tools disagree, compare definitions before assuming an error. Their question sets, weights and comparison groups may differ even when the labels look alike.
What should you avoid concluding?
A 20% AI share does not mean that 20% of potential customers know the business. Nor does it establish the same share of revenue or website traffic. AI visibility and actual customer behavior require different evidence.
Your share can rise because competitors appear less often. Examine the absolute numerator too. An unchanged count calls for a different interpretation from newly appearing in more relevant questions.
How can you use the comparison?
Maintain a stable comparison group and disclose added or removed brands. Break down important topics without selecting only favorable results. Explain why the prompt set reflects real business questions.
Then return to individual answers. Which sources provide useful information? Where is your own explanation incomplete? Which gap matters to customers? The percentage identifies an area to investigate; the actual content and circumstances determine the work worth doing.