AI Share of Voice: how to track what ChatGPT, Perplexity and Gemini say about your brand
AI Share of Voice is the measure of how often and how prominently your brand is mentioned across AI-generated answers, compared with your competitors. It tracks your presence inside the answers ChatGPT, Perplexity, Gemini and Copilot give directly to buyers, and it is replacing rankings as the headline KPI for AI search.
For two decades the question was “where do we rank?” In 2026, with AI assistants answering questions directly, a sharper question has taken over: when a customer asks ChatGPT for the best option in your category, does your name come up, and what does the AI say about you?
TL;DR: Quick Answer
AI Share of Voice is how often and how prominently AI assistants name your brand in their answers, compared with competitors. Where traditional share of voice tracked visibility across search results or ads, AI Share of Voice tracks your presence inside the answers ChatGPT, Perplexity, Gemini and Copilot give directly to buyers. Because an AI answer typically names only three to five brands, it is replacing rankings as the headline KPI for generative engine optimisation. Run a three-step audit, a Prompt Audit, a Source Audit and a Descriptor Audit, to learn whether you are mentioned, why, and how AI describes you.
Key takeaways
- AI answers name only three to five brands, so you are either in the shortlist or invisible: there is no page two in an AI answer
- LLMs form a consensus from five to ten third-party sources (the Multi-Vote concept), so you cannot simply optimise your own page and expect to be named
- A three-step audit, Prompt, Source and Descriptor, tells you whether you are mentioned, why, and exactly how AI describes you
- Win a specific, ownable use-case descriptor rather than chasing “best overall”, which a small SA brand will never reliably win
- Track share of voice, share of model and descriptor consistency, and weight descriptor consistency most heavily
- Expect descriptor shifts within roughly 60 to 90 days: AI Share of Voice moves slower than rankings but compounds once earned
For two decades, the question was “where do we rank?” In 2026, with AI assistants answering questions directly, a sharper question has taken over: when a customer asks ChatGPT for the best option in your category, does your name come up, and what does the AI say about you? That is AI Share of Voice, and it is becoming the metric that matters.
What is AI Share of Voice?
AI Share of Voice is the measure of how often and how prominently your brand is mentioned across AI-generated answers, compared with your competitors. Where traditional share of voice tracked your visibility across search results or ads, AI Share of Voice tracks your presence inside the answers ChatGPT, Perplexity, Gemini and Copilot give directly to buyers. It is your slice of the AI conversation.
The shift matters because AI answers are far more selective than a search results page. A Google results page shows ten blue links and gives a long-tail brand a fighting chance. An AI answer typically names only three to five brands. You are either in that shortlist or you are invisible. There is no page two in an AI answer.
This is why AI Share of Voice is replacing rankings as the headline KPI for generative engine optimisation. With around 58 to 60% of searches now ending without a click, and most users now leaning on AI summaries a meaningful share of the time, the AI answer increasingly is the destination. Being ranked tenth on Google means little if the AI answer above the results never mentions you.
Why does AI Share of Voice replace keyword rankings?
Because AI answers compress ten options into three to five, and rankings no longer describe how customers actually find you. A position-six ranking might still earn clicks in classic search, but if the AI Overview or ChatGPT answer above it names only your three competitors, you have lost the customer before they ever scroll. AI Share of Voice measures the outcome that now decides the sale.
There is a mechanism behind this worth understanding: the Multi-Vote concept. LLMs do not simply read your website and decide you are great. They form a consensus by synthesising five to ten third-party sources, each effectively a “vote” for what your brand is and whether it belongs in the answer. A review on a comparison site, a Reddit thread, a directory listing, an industry roundup, each contributes a vote. The brands with the most consistent, credible votes win the mention.
This reframes the whole game. You cannot simply optimise your own page and expect to be named. You have to earn a distributed reputation across the sources the AI reads. It also explains why brands with both a citation and a mention tend to resurface more often in future answers; consistency across sources compounds.
How do you run an AI Share of Voice audit?
Run a three-step audit: a Prompt Audit to see what AI says when buyers ask, a Source Audit to find which sources are voting, and a Descriptor Audit to learn the exact words AI uses to describe you. Together they tell you whether you are mentioned, why, and how, which is everything you need to improve your standing. Here is the practical method.
| Audit Step | What You Do | What It Reveals |
|---|---|---|
| 1. Prompt Audit | Run 10-20 buyer-intent prompts through ChatGPT, Perplexity and Gemini | Whether you appear, in what position, and which competitors appear alongside you |
| 2. Source Audit | Log every source each AI cites, then tally votes by domain and source type | Which sites the AI trusts, where competitors dominate, and where you are absent |
| 3. Descriptor Audit | Extract the exact, repeated phrases the AI uses to describe your brand | Your AI reputation in plain words, so you can shape it deliberately |
Step 1: The Prompt Audit
List 10 to 20 real, buyer-intent prompts a South African customer would actually type, then run each through ChatGPT, Perplexity and Gemini and record who gets named. Focus on prompts that signal purchase intent rather than vanity searches for your own name. Strong prompt patterns include:
“best [your category] in South Africa”“[competitor] vs [your brand]”“[your brand] alternatives”“affordable [category] for small business in Johannesburg”“is [your brand] any good”
Log, for each prompt and each platform, whether you appear, in what position, and which competitors appear alongside you.
Step 2: The Source Audit
For every AI answer, log each source the AI cites or draws on, then tally the votes by domain and by source type. You are looking for patterns: which sites the AI trusts in your category, which competitors dominate those sources, and where you are absent. Worth knowing: studies of AI Overview citations find that around half come from pages already ranking in classic search, so traditional SEO and AI visibility reinforce each other.
Group the votes by type, your own site, third-party reviews, directories, editorial roundups, community threads like Reddit, so you can see where your reputation is thin and which source types you most need to win.
Step 3: The Descriptor Audit
Extract the exact, repeated phrases the AI uses to describe your brand. This is the most overlooked and most valuable step. If ChatGPT consistently calls you “a budget-friendly option for startups,” that descriptor is your AI reputation, whether you chose it or not. Your job is to identify it, then shape it deliberately.
The strategic rule: win a specific, ownable use-case descriptor rather than chasing “best overall.” No AI will reliably crown a small SA agency “best overall” against global names, and trying is wasted effort. But “the specialist for POPIA-compliant e-commerce in South Africa” is winnable, and far more valuable because it matches high-intent prompts.
“The brands that win AI Share of Voice are not the ones shouting ‘best overall’. They are the ones that own a specific, ownable descriptor across the sources the AI reads. Get the descriptor right and the mentions follow, because the model keeps describing you the same way to every buyer who asks.”
Cobus van der Westhuizen, CEO & Digital Strategist, Juicy Designs, reviewed and verified March 2026
An AI Share of Voice audit has three steps: a Prompt Audit, a Source Audit and a Descriptor Audit. Run 10 to 20 buyer-intent prompts across ChatGPT, Perplexity and Gemini, log every source each answer cites, then extract the exact phrases the AI repeats about your brand. The strategic goal is to win a specific, ownable use-case descriptor rather than “best overall”. Source: Juicy Designs generative engine optimisation methodology, South Africa, 2026.
The metrics to track for AI Share of Voice
Track three: share of voice (how often you are named versus competitors across all prompts), share of model (how your presence differs across ChatGPT, Perplexity and Gemini, since each weights sources differently), and descriptor consistency (whether the AI describes you the same way across platforms). Prioritise descriptor consistency over raw mention volume.
Brands a typical AI answer names when a buyer asks for the best option in a category. With no page two in an AI answer, you are either in that shortlist or invisible, which is why mention volume alone is a vanity metric.
Source: Juicy Designs AI search analysis, 2026The reason to weight descriptor consistency so heavily is that a consistent, accurate, ownable descriptor is what actually converts. Being mentioned a lot but described vaguely, or worse, inaccurately, does not win business. Being reliably described as the right brand for a specific need does. Volume is a vanity metric; descriptor control is the real outcome.
For tooling, you can use dedicated platforms such as Profound, Peec AI or Otterly, or run the whole audit in a spreadsheet to start. A spreadsheet is genuinely enough to begin; do not let tool selection delay the first audit. On timeline, expect descriptor shifts within roughly 60 to 90 days once you start improving the sources that feed the AI’s consensus. AI Share of Voice moves slower than rankings but compounds. For a managed approach, see how we measure AI search readiness and how to get cited by ChatGPT, Perplexity and AI Overviews.
Track three AI Share of Voice metrics: share of voice, share of model and descriptor consistency. Share of voice is how often you are named versus competitors; share of model is how your presence differs across ChatGPT, Perplexity and Gemini; descriptor consistency is whether the AI describes you the same way across platforms. Weight descriptor consistency most heavily, because a consistent, ownable descriptor converts where raw mention volume does not. Source: Juicy Designs, South Africa, 2026.
Frequently asked questions
How is AI Share of Voice different from normal share of voice?
Traditional share of voice measures your visibility across search results, ads or media coverage. AI Share of Voice measures how often and how prominently AI assistants name your brand inside their direct answers, compared with competitors. The key difference is selectivity: AI answers name only three to five brands, so the threshold for visibility is much higher.
How many prompts do I need to audit my AI Share of Voice?
Start with 10 to 20 genuine buyer-intent prompts covering your category, competitor comparisons and alternatives, run across ChatGPT, Perplexity and Gemini. That is enough to reveal clear patterns in who gets mentioned and how you are described. You can expand later, but a focused set of high-intent prompts gives reliable signal faster than a huge unfocused list.
How long until I see my AI Share of Voice improve?
Expect roughly 60 to 90 days for descriptor and mention shifts, because AI answers draw on a consensus of third-party sources that must update first. Improving those sources, such as reviews, directories and editorial mentions, takes time to propagate into the model's responses. AI Share of Voice moves more slowly than keyword rankings but is far stickier once earned.