Every organic programme tracks keyword rankings. Most now track AI citations too. But the question that neither metric answers on its own is: when someone asks an AI assistant about your category, how often does your brand appear compared to your competitors?

That is what AI share of voice measures. It is the competitive metric that sits on top of citation tracking and tells you whether your brand is gaining or losing ground in AI-generated answers. Without it, you know whether you appeared. With it, you know whether you are winning.

What AI Share Of Voice Actually Measures

AI share of voice is the percentage of brand mentions your company receives across AI-generated responses, relative to all brand mentions for your category on the same set of prompts. It is a competitive metric by definition. A citation count tells you about your own performance. AI search share of voice tells you about your position relative to named competitors.

The concept is borrowed from traditional marketing, where share of voice historically meant your brand's share of total advertising impressions in a market. In AI search, the "impressions" are brand mentions inside AI-generated answers, and the market is defined by a tracked set of prompts that represent your category.

The Calculation

The formula is straightforward.

AI Share of Voice = (your brand mentions / total brand mentions across tracked prompts) x 100

If your tracking covers 100 prompts across ChatGPT, Perplexity, Gemini, and Google AI Overviews, and those prompts generate 300 total brand mentions across all competitors, and your brand appears 60 times, your SOV is 20%.

The number only means something in context. Twenty percent in a fragmented market with fifteen competitors may represent category leadership. Twenty percent in a two-player market means you are losing.

Why The Number Varies By Platform

One of the most important properties of this metric is that it is not uniform across AI engines. Your LLM share of voice on ChatGPT may be materially different from your share on Perplexity or Google AI Overviews, because each platform retrieves from different source pools, weights authority differently, and serves different query types.

ChatGPT tends to draw from broadly authoritative web content and reflects patterns from its training data. Perplexity runs real-time web retrieval and surfaces recent content more aggressively. Google AI Overviews are closely tied to organic ranking signals. Google AI Mode leans toward transactional and conversational queries.

This means a brand can hold strong geo share of voice on Google AI Overviews, where its organic SEO foundation gives it an advantage, while barely registering on Perplexity, where it has no recent editorial coverage. 

Our research on ​how brands show up differently across AI platforms documents exactly how much citation patterns diverge across engines for the same queries. Tracking a single platform and treating it as the full picture is one of the most common measurement errors in AI visibility right now.

How To Set Up Tracking That Produces A Useful Number

The quality of your SOV data depends entirely on two inputs: the prompts you track and the competitors you include in the comparison set.

Prompt Selection

The prompts need to represent the queries your buyers actually ask AI tools. This is not the same as your SEO keyword list. AI prompts are typically longer, more conversational, and framed as questions or requests for recommendations.

Start with 15 to 30 prompts that cover your core category. Segment them by intent: awareness prompts ("what is X"), consideration prompts ("best X for Y"), and decision prompts ("X vs Y"). This segmentation matters because your share of voice may be strong at the top of the funnel and absent at the bottom, or the reverse.

Competitor Set

Define three to five named competitors. The total brand mentions in the denominator should include only the brands you are actually competing against, not every brand the AI has ever mentioned. A tightly scoped competitor set produces a number you can act on. An open-ended one produces noise.

Measurement Cadence

AI answers change. The same prompt can produce different brand mentions on different days, because the models update their retrieval and generation behaviour regularly. Monthly tracking is the minimum useful cadence. Weekly tracking is better for brands actively investing in GEO and wanting to see whether specific content changes are moving the number. Passionfruit tracks this through ​AI visibility tracking, running prompt sets across all major platforms on a recurring schedule tied to the content and citation work driving the numbers.

What Good Looks Like

There is no universal benchmark, because the number is category-specific and competitor-set-dependent. But patterns are emerging.

In fragmented markets with ten or more competitors, 15 to 20% SOV typically represents a leadership position. In consolidated markets with two to three dominant players, anything below 30% means you are losing ground to whoever is above you. The more useful signal is directional momentum: a brand climbing from 8% to 15% over 60 days is in a stronger position than a brand flat at 25% while a competitor climbs behind it.

Position Within The Response Matters Too

Raw mention frequency is not the whole picture. Where your brand appears within an AI-generated response, first in the list or last, affects both perception and click-through. A brand mentioned first in a recommendation answer carries more weight than one mentioned as an afterthought at the end. Some tracking tools now surface average position alongside mention frequency, which gives a more complete picture of competitive standing.

Sentiment Adds A Second Dimension

A high share of voice with negative sentiment is worse than a lower share with positive framing. If your brand appears frequently but only in "alternatives to" lists or negative comparisons, the mention volume is not an asset. Tracking sentiment per mention catches framing problems that a raw mention count would miss. Our research on ​whether citations are the best visibility metric to track examines the broader question of which AI visibility signals actually correlate with business outcomes.

What Moves The Number

Improving AI search share of voice requires several workstreams working together. Content that directly answers tracked prompts, earned media that strengthens third-party visibility, and consistent brand-to-topic associations can all increase the likelihood of being mentioned.

Results vary by category and platform, so the goal is to track these changes against your prompt set over time rather than expect an immediate shift. Passionfruit's GEO service connects these workstreams to prompt-level share of voice data, giving teams a measurable view of competitive AI visibility. Talk to the team about where your brand stands in AI answers and what could move the number.

Frequently Asked Questions

How Is AI Share Of Voice Different From Citation Count?

Citation count measures how often your brand is mentioned in AI answers. Share of voice adds competitive context by expressing your mentions as a percentage of total brand mentions for your category. You can have a high citation count and still be losing share if competitors are growing faster.

How Many Prompts Should I Track To Get A Reliable Share Of Voice Number?

Start with 15 to 30 prompts covering your core category, segmented by awareness, consideration, and decision intent. Fewer than 15 produces too volatile a number. More than 50 is useful for mature programmes, but not necessary for establishing a baseline.

Does Improving Google Rankings Automatically Improve AI Share Of Voice?

Partially. Google AI Overviews pull heavily from organic ranking signals, so strong SEO helps on that surface. But ChatGPT and Perplexity use different retrieval logic, and improving organic rankings does not guarantee increased mentions on those platforms.

How Often Do AI Answers Change For The Same Prompt?

Frequently. The same prompt can produce different brand mentions on different days as models update their retrieval behaviour. This is why single-point-in-time checks are unreliable and recurring measurements on at least a monthly cadence are necessary.

Can A Small Brand Compete On AI Share Of Voice Against Larger Competitors?

Yes, particularly on narrow, specific prompts where the larger competitor has less targeted content. AI engines retrieve the best answer for a given query, not the biggest brand. A smaller brand with more specific, well-structured content for a niche prompt set can hold a disproportionate share of voice in that segment.