AI Share of Voice: How to Calculate It in ChatGPT, Gemini and Perplexity (2026)

AI Share of Voice is the share of brand mentions inside a set of AI-generated answers that goes to your brand. The denominator is the total relevant brand mentions in the answers you monitor: your brand plus the competitors you track, in the prompts, models, languages and markets you defined.

A buyer can ask ChatGPT which software to choose, Gemini which provider to compare, Perplexity which sources support a recommendation or Google AI Overviews which options exist in a category. The answer shows a compressed selection of brands, arguments and sources. If your brand is absent there, it is not among the options that answer presents.

The metric answers one concrete question: when someone asks for a recommendation in your category, what share of the mentions goes to your brand?

This guide explains what AI Share of Voice is, how to calculate it, how to weight mentions if your use case needs it, which mistakes to avoid and how to use it to prioritize content, SEO, PR and product work.

What AI Share of Voice is

AI Share of Voice is a brand's share of presence inside answers generated by models such as ChatGPT, Gemini, Perplexity, Claude or Google AI Overviews, compared with the presence of its competitors.

It does not only measure whether a brand appears. It measures how much it appears relative to everyone else.

Imagine you monitor 100 high-intent questions for your category. Across those answers, the models mention 220 brands in total: your brand appears 28 times, competitor A 55 times, competitor B 42 times and other brands make up the rest. Your basic Share of Voice would be 28 / 220 = 12.7%.

That calculation is a useful first read, but it is limited. Not all mentions are equal. Being the first recommendation in a buying question is not the same as being a secondary example in an educational answer. A Perplexity citation, an explicit ChatGPT recommendation and a source link in AI Overviews are also different signals.

That is why it helps to combine two layers:

• Basic Share of Voice: how often you appear relative to the total brand mentions.
• Weighted Share of Voice: how much each mention counts based on intent, position, prominence, model, market and quality.

If you are still building the full KPI framework, start with the GEO metrics you should track. This article goes into detail on one of them.

The basic AI Share of Voice formula

The basic formula is simple:

AI Share of Voice = your brand mentions / total relevant brand mentions x 100

Example:

• Your brand: 36 mentions
• Competitor A: 52 mentions
• Competitor B: 41 mentions
• Competitor C: 21 mentions
• Other competitors: 30 mentions
• Total: 180 mentions

Your basic Share of Voice would be 36 / 180 = 20%.

This number gives an initial picture of the conversation. If your brand has 20% and the main competitor has 28.9% (52 / 180), there is a competitive gap. If your share rises from 12% to 20% with the same prompt set, models and configuration, that is an observed change; before treating it as real, check that it exceeds the normal variation between repeated runs and that the number of mentions is large enough.

The basic formula has three limits:

1. It treats every question equally. A mention in "what is GEO" counts the same as a mention in "best tools to measure AI visibility", even though the second query has much more commercial intent.

2. It treats every position equally. Appearing first, third or as a secondary source adds one unit, which may not reflect what you want to measure.

3. It does not measure quality. A positive mention, a neutral comparison and an incorrect description count the same unless you classify them.

The basic formula is a good starting point. If you need to reflect these differences, add a documented weighting.

The advanced formula: weighted Share of Voice

Weighted Share of Voice assigns a different value to each appearance. A practical way to calculate it:

Mention score = intent weight x prominence weight x model weight x quality factor

Then:

Weighted Share of Voice = your brand score / total score for all brands x 100

The weights are internal decisions, not validated values: document them, keep them fixed during the comparison period and always report the basic version alongside.

Intent weight

Not all prompts have the same value. You can use a simple scale:

• Informational: 1 point
• Comparative: 2 points
• Commercial or decision: 3 points
• Local or transactional: 3 points
• Critical reputation: 2 or 3 points, depending on risk

A question like "what is AI Share of Voice" supports authority. A question like "which tool do you recommend to measure Share of Voice in ChatGPT and Gemini" is closer to a purchase. If that difference matters for your decisions, you can give them different weights; leaving them equal is also a valid choice as long as you state it.

Prominence weight

An example scale:

• First recommendation or highlighted brand: 3 points
• Main list, not first: 2 points
• Secondary mention or example: 1 point
• Cited source without a clear recommendation: 0.5 to 1 point
• Incorrect or negative mention: can be tracked separately as risk

This layer connects with why position in AI answers matters. Besides being in the answer, where and how you appear matters.

Model weight

For some brands ChatGPT will matter most; for others Perplexity, Gemini or AI Overviews will have more impact. Start with the same weight for every model and adjust later based on audience, market and conversions.

Example:

• ChatGPT: 1.0
• Gemini: 1.0
• Perplexity: 1.0
• Google AI Overviews: 1.2 if your data show your category depends heavily on search
• Claude: 0.8 if your data show your audience uses it less

The key is documenting the logic. If weights change without explanation, the dashboard stops being comparable.

Quality factor

Mention quality can adjust the result:

• Positive and accurate recommendation: 1.0
• Neutral mention: 0.7
• Secondary mention with little actionability: 0.5
• Incorrect information: 0 or a risk tag
• Negative mention: separate as reputation risk

The quality factor prevents counting as positive a mention that does not help you. If a model mentions your brand but describes it as a budget option when you sell to enterprises, that visibility may attract the wrong buyer.

Which prompts to use for Share of Voice

The prompt set determines how useful the metric is. If the questions are too generic, your Share of Voice will say little about real demand. If you only use branded questions, you will not see the category.

A good set mixes six types:

Category prompts. "Best AI visibility tools", "platforms to measure presence in ChatGPT", "GEO software for marketing teams".

Problem prompts. "How to know if AI recommends my brand", "how to measure whether my company appears in Gemini", "why ChatGPT recommends my competitors".

Comparison prompts. "Alternatives to [competitor]", "[brand] vs [competitor]", "best options for agencies managing several clients".

Decision prompts. "Which tool do you recommend to measure AI Share of Voice", "which platform would a CMO use to report visibility in ChatGPT and Perplexity".

Market prompts. "AI visibility tools for companies in the United States", "GEO software for brands in the UK", "recommended providers for B2B SaaS".

Reputation prompts. "Is [brand] reliable", "[brand] reviews", "which brands lead [category]".

For each prompt, store language, country, model, intent, date, brands mentioned, position, sources and the full response. Without that structure you cannot explain where the number comes from.

For a broader view of presence auditing, see how to measure brand visibility in AI and the AI visibility tracker.

How to interpret the result

AI Share of Voice should be read by segment, not only as one global number.

By model. You might have 32% in Perplexity and 8% in Gemini. That difference is worth investigating: check which sources each model cites and how your brand is described across Google's ecosystem.

By intent. You might have 40% in educational questions and 5% in buying questions. That indicates you appear in informational questions but rarely when a recommendation is requested.

By market. Your brand can appear often in English and rarely in Spanish, or do well in the US and poorly in the UK. A global average would hide the issue.

By competitor. Knowing you are losing share is not enough. You need to know to whom and on which prompts. One competitor may dominate comparisons, another local prompts and another reputation questions.

By source. In Perplexity and AI Overviews, cited sources give clues about the share. If competitors appear because directories or comparison pages recommend them, the action is not only publishing more blog posts; it is working on external presence.

By quality. More mentions are not always positive. If a model starts describing you incorrectly or with outdated positioning, share can rise while the quality of the description gets worse.

What to do if your Share of Voice is low

Low Share of Voice is not fixed by publishing random content. First understand why others appear.

If competitors win because they have clearer use-case pages, create or improve those pages. If they win because they appear in external comparisons, work on PR, partners, directories and reviews. If they win in AI Overviews, analyze the sources and content Google cites. If they win in Gemini, review how your brand is described, the consistency of your site and your profiles on Google services. If they win in Perplexity, review citable sources, concrete data and content that answers directly.

Recommended sequence:

1. Define a stable prompt set.
2. Calculate basic Share of Voice, and the weighted version if you need it.
3. Segment results by model, intent and market.
4. Identify the competitors and sources explaining the gap.
5. Prioritize content, schema, PR, profile and product actions.
6. Measure again with the same set.

If you change the prompt set every week, you will not know whether you improved or only changed the sample.

How to connect Share of Voice with ROI

AI Share of Voice is not ROI by itself. It is a competitive presence metric. To connect it with business outcomes, cross it with demand indicators.

Cross share by cluster with:

• Search Console impressions and clicks.
• Branded searches.
• Direct and organic traffic by market.
• Leads by source and landing page.
• Assisted conversions.
• Pipeline or sales by category.
• Changes in the questions sales teams hear.

If your Share of Voice rises in commercial prompts while branded searches or leads in related categories grow, you have a coincidence worth following, although it does not show that one caused the other. If share rises in educational prompts and nothing changes in the business indicators, there is no evidence of impact.

To defend budget, combine this metric with the methodology for GEO ROI and attribution. Share of Voice describes your presence relative to competitors; ROI requires separate business data.

Common mistakes when measuring AI Share of Voice

Reading the basic count as if it reflected intent or prominence. An unweighted count is a valid metric; the mistake is presenting it as something it does not measure. If you need those distinctions, add a documented weighting.

Measuring only one model. Other models can give different results; choose the ones your audience uses.

Not saving the full response. Without the full text you cannot audit framing, errors, sources or changes.

Mixing languages and countries. Models can recommend different brands by market. Do not mix them if you sell regionally.

Ignoring emerging competitors. Sometimes the risk is not the obvious leader but a new brand starting to appear in buying prompts.

Not separating recommended brand and cited source. In AI Overviews and Perplexity, a source can be cited without the brand being recommended. These are different signals.

Changing the prompt set constantly. A stable dashboard needs a comparable baseline.

Counting incorrect mentions as positive. If a model describes you incorrectly, it is a risk that needs correcting.

Pair this with how to measure brand mentions in Google AI Overviews to also cover the search ecosystem.

How Mentio supports the calculation

Mentio generates questions by market and language, runs them on the models included in your plan and detects your brand and your competitors, with position and tone. From those mentions you can calculate basic Share of Voice with the formula in this article; the weighting, if you use it, is a decision you document yourself. Mentio does not measure Google AI Overviews, so that surface needs separate tracking.

The useful reading is the one that leads to actions: if a competitor appears because of external comparisons, work on those sources; if Perplexity cites pages where you are absent, work on those pages; if you appear with the wrong description, correct your positioning.

AI Share of Voice checklist

• Define real competitors by category and market.
• Build a prompt set by intent, not only by keyword.
• Measure the models your audience uses; add AI Overviews when search is important.
• Save the full response, date, country, language and model.
• Classify mentions by position, prominence and quality.
• Calculate basic Share of Voice; add the weighted version only if your use case needs it.
• Segment by model, intent, market and competitor.
• Review cited sources when they exist.
• Connect changes with branded search, leads and pipeline.
• Keep the prompt set stable to compare trends.

AI Share of Voice does not replace rankings, traffic or conversions. It shows how the mentions in your category are distributed between your brand and competitors within the measured set of answers. It is not commercial market share.

Before reading a share as market position, confirm the cohort is valid: the AI visibility industry benchmark explains how to define the peer population and comparable cells behind that percentage.

Calculating the indicator and setting the weighting rule are separate decisions: see how to weight a prompt bank by demand and business value before applying a weighted formula.

Frequently asked questions

What is AI Share of Voice?

It is a brand's share of presence inside AI-generated answers compared with its competitors. It covers appearances, prominence, models, markets and mention quality.

How do you calculate AI Share of Voice?

The basic version divides your brand mentions by all relevant brand mentions in the answers you monitor. An optional weighted version multiplies each mention by factors you define for intent, position, model, market and quality.

Is it the same as mention rate?

No. Mention rate measures the percentage of questions in which your brand appears. Share of Voice compares your presence with that of your competitors.

Which models should I measure?

The ones your audience uses. ChatGPT, Gemini and Perplexity are a common starting point; depending on your market, also Claude, Google AI Overviews or Grok.

How many prompts do I need?

There is no universal number. It depends on how many intents and markets you want to cover and how much precision you need: with few mentions, small variations change the percentage a lot. Start with a set covering your main intents and expand it by country, language or segment.

How often should I update it?

Choose a fixed cadence you can sustain, for example monthly, so periods are comparable. During launches or reputation issues it can make sense to increase it.

Does Mentio measure AI Share of Voice?

Mentio measures mentions of your brand and your competitors in the models included in your plan. With that data you can calculate basic Share of Voice; on Pro and Enterprise plans it also generates recommendations.

Want to know if AI mentions your brand?

Discover your visibility in ChatGPT, Claude and Gemini in minutes.

Analyze your brand for free

Related articles