Photograph by Orhan Pergel via Pexels. Illustrative image, not a customer or endorsement.
Blog · · 6 min read · NEXUM Team
AI Share of Voice: How to Measure Your Brand's Visibility in AI Search
Short answer: AI share of voice is the percentage of relevant AI answers that mention your brand compared with your competitors. To measure it, pick a fixed set of realistic buyer prompts, run them repeatedly across the AI assistants your buyers use (ChatGPT, Gemini, Perplexity, Claude), record which brands appear and how they are described, and calculate your share over time. Single checks are misleading; trends over weeks are what count.
This guide explains the metric, the pitfalls, and a step-by-step method you can start with today.
Why traditional metrics miss AI visibility
Rank tracking tells you where a URL sits on a results page. Organic sessions tell you how many people clicked. Neither tells you whether an AI assistant recommended you before anyone clicked anything.
That gap is growing. On Google, AI features now reach massive audiences: Google said in 2026 that AI Overviews has over 2.5 billion monthly active users and AI Mode has passed 1 billion. Outside Google, ChatGPT reaches 1.2 billion weekly users, according to OpenAI.
Meanwhile, when an AI Overview appears, Ahrefs found the top organic result gets a 58% lower click-through rate. Many buyers form their shortlist inside the answer. If you are not measuring that answer, you are missing part of the funnel.
What AI share of voice actually measures
A good AI share of voice framework separates a few related ideas.
Mentions vs citations
- A mention is when the answer names your brand.
- A citation is when the answer links to a URL as a source.
They are not the same. An answer can recommend your brand while citing a third-party review. STAT's Tom Capper summarized it in a Search Engine Journal webinar recap: citations have URLs, but mentions have entities. For most brands, being mentioned and recommended is the outcome; citations help explain why.
Presence vs recommendation
Being listed fifth with a caveat is different from being the first suggestion. Track position and context, not only presence.
Sentiment and accuracy
Is the AI describing you correctly? Wrong pricing, outdated features or a confused positioning can hurt more than absence.
The core formula
For a given set of prompts, engines and time period:
AI share of voice = (answers that mention your brand ÷ total answers sampled) compared with the same rate for each competitor
A simple way to report it:
| Brand | Answers mentioning brand | Mention rate | Share of all brand mentions |
|---|---|---|---|
| Your brand | A | A ÷ total answers | A ÷ (A + B + C) |
| Competitor 1 | B | B ÷ total answers | B ÷ (A + B + C) |
| Competitor 2 | C | C ÷ total answers | C ÷ (A + B + C) |
Illustrative structure only; plug in your own sampled data.
Mention rate tells you how visible you are on your own. Share of mentions tells you how you compare.
Why one check is never enough
AI answers are not fixed. Ask the same question twice and you may get different brands. Ahrefs tracked 37,500 citations and found that 50–83% of citations are gone the next day, but about 75% of those return within 20 days. Citations that lasted eight or more days almost always came back after a drop.
Ahrefs draws three lessons that apply directly to measurement:
- Count frequency over a window, for example how many of the last seven days you were mentioned, instead of reacting to one result.
- Brand mentions are a more stable KPI than individual citations.
- Separate prompt groups. Questions about your own brand (pricing, reviews) behave differently from discovery prompts ("best X for Y").
Treat AI visibility like a sampling problem. The more prompts and repeated runs you have, the more trustworthy your trend line.
How to measure AI share of voice, step by step
Step 1: Define your competitive set
Choose three to eight competitors that your buyers actually compare you with. Include sub-brands or product names if buyers use them.
Step 2: Build a realistic prompt set
Avoid pasting a keyword list into a chatbot. Real users ask with context. Start from:
- Discovery prompts: "What's the best [category] for [use case]?"
- Comparison prompts: "[Your brand] vs [competitor]"
- Alternative prompts: "Alternatives to [competitor]"
- Brand prompts: "Is [your brand] worth it?" "How much does [your brand] cost?"
Add personas, budgets or locations where they change the answer. Aim for at least 20–50 prompts to start, then grow.
Step 3: Choose engines and a schedule
Cover the assistants your buyers use. For most B2B and SaaS companies that means ChatGPT, Gemini, Perplexity and Claude, plus Google's AI Overviews and AI Mode where relevant. Run the full set on a fixed schedule (daily or several times a week) so you can see trends.
Step 4: Record the right fields
For each answer, capture: engine, prompt, date, brands mentioned, order of mention, whether you were recommended, how you were described, and which URLs were cited.
Step 5: Calculate and segment
Calculate mention rate and share of voice overall, then split by engine, prompt group and competitor. A brand can lead in ChatGPT and be absent in Perplexity, or win brand questions but lose discovery questions.
Step 6: Investigate the sources
When a competitor wins, look at what the engine cited. Research from Profound, covering over 4 billion AI citations, found Reddit was the most-cited domain across major answer engines, followed by YouTube and Wikipedia. Your category will have its own mix of review sites, publications and communities. Those sources are where you can act.
Step 7: Connect visibility to outcomes
Share of voice is a leading indicator. Pair it with:
- AI referral traffic and conversions in GA4.
- Google's Generative AI performance report in Search Console, which Google recommends in its official AI optimization guide.
- A "how did you hear about us?" question at signup, since many AI-influenced visits arrive later as direct or branded traffic.
Common mistakes
- Tracking too few prompts. A handful of prompts produces noisy, misleading numbers.
- Reacting to daily swings. Volatility is normal; look at weekly or monthly trends.
- Only counting citations. You will miss recommendations that link elsewhere.
- Measuring one engine. Each assistant retrieves and weighs sources differently.
- Ignoring accuracy. A mention with wrong pricing or features is a problem to fix, not a win.
How NEXUM measures AI share of voice
NEXUM is a SaaS platform built specifically for this. You set your brand, competitors and prompts, and NEXUM tracks mentions, positioning and cited sources across ChatGPT, Gemini, Perplexity and Claude (Copilot and Grok are coming soon). It shows your share of voice by engine and over time, highlights the sources behind competitor wins, and connects visibility to traffic through its GA4 integration. When you find a gap, the Content Generator helps you create content aimed at it.
Plans range from the free Explorer plan (one AI engine) to Lite ($79/month), Pro ($179/month), Growth and Advanced ($699/month), and every paid plan has a 7-day free trial. Measure your AI share of voice at hellonexum.com.
FAQ
What is AI share of voice?
It is the proportion of relevant AI-generated answers that mention your brand, compared with your competitors, across a defined set of prompts and AI engines over a period of time.
How many prompts do I need?
Enough to cover your main buying situations. Start with 20–50 realistic prompts across discovery, comparison, alternatives and brand questions, and expand as you learn which ones matter.
How often should I measure?
Regularly enough to see trends despite daily volatility. Daily or several times per week, reported as weekly or monthly averages, works well for most brands.
Is a citation or a mention more important?
For most brands the mention or recommendation is the outcome that matters. Citations are useful to understand which sources influence the answer.
Can I measure AI share of voice manually?
Yes, with a spreadsheet and patience. It becomes hard to sustain once you track several engines, competitors and dozens of prompts, which is where a dedicated tool saves time.
