AI Share of Voice: What It Is and How to Measure It

AI share of voice is the share of relevant AI answers that mention your business compared with your competitors. You measure it by asking AI assistants the questions your customers ask, recording which brands each answer names, and calculating how often you appear relative to everyone else.

It is the AI-search version of a familiar marketing metric. Instead of counting ads or search rankings, you count mentions inside answers from ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews and Copilot.

What does AI share of voice actually measure?

When someone asks an AI assistant “What is the best accounting software for freelancers?” or “Who is a good family dentist in Dubai Marina?”, the answer usually names a handful of businesses. Some brands show up again and again. Others never appear.

AI share of voice turns that pattern into something you can track. It answers two practical questions:

• How often am I named when people ask questions I should be part of?
• How does that compare with the competitors who are named instead?

There are two closely related numbers, and it helps to keep them separate:

• Mention rate: the share of answers that name you at all. This is about you alone.
• Share of voice: your mentions as a share of all brand mentions across those answers. This is about you versus the field.

Mention rate tells you whether you are visible. Share of voice tells you whether you are winning the conversation. You want both.

How do you calculate AI share of voice?

The simplest formulas look like this:

• Mention rate = answers that mention you ÷ total answers checked
• Share of voice = your brand mentions ÷ total brand mentions (you plus all competitors) across the same answers

Count one mention per brand per answer. If an answer names your company three times, it still counts once. This stops long answers about a single brand from distorting the picture.

A worked example (hypothetical numbers)

The following is an illustration with made-up numbers, not real data. Imagine a physiotherapy clinic in London. The owner writes 20 questions customers might ask, such as “best sports physio in central London” and “where can I get physio for a running injury near Shoreditch”. She asks each question once in one AI assistant and records the brands named.

In this example:

• Her mention rate is 5 ÷ 20 = 25%. She appears in a quarter of the answers.
• Total brand mentions are 5 + 12 + 8 + 5 = 30. Her share of voice is 5 ÷ 30, roughly 17%.
• Competitor A’s share of voice is 12 ÷ 30 = 40%. That is the brand to study.

Again, these figures are invented to show the arithmetic. Your own numbers will depend entirely on your questions, market and engines.

Should position in the answer count?

Being named first in a list is not the same as being named fifth. Some teams add a simple weighting, for example giving more credit to the first brand mentioned. That can be useful, but keep it optional. Start with plain mention counts, which are easier to explain to a client or a manager, and record the order separately so you can look at it later.

How many questions do you need for a reliable number?

This is where most share-of-voice reports go wrong. AI answers vary by engine, wording, location, time and personalisation. Ask the same question twice and you may get a different list of brands.

A few cautions:

• Small samples swing wildly. With only a handful of answers, one extra mention can move your percentage a lot. Treat early numbers as a rough signal, not a verdict.
• Repeat the same questions over time. A trend across several checks tells you far more than a single snapshot.
• Keep the question set stable. If you change half your questions between checks, you cannot compare the results.
• Choose questions with real intent. “Best wedding photographer in Sydney” matters more than a vague question nobody types.
• Note the method. API answers can differ from what people see in consumer apps, so record how each answer was collected.

The honest goal is not a perfectly precise number. It is a consistent measurement you can repeat, so you can see whether things are moving in the right direction. For more on this, see how to measure AI visibility improvements.

Why break share of voice down by engine and location?

A single blended number hides useful detail. Each AI assistant searches the web differently, uses different sources and may favour different brands.

Per-engine breakdown

Calculate share of voice separately for each engine you care about: ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews and AI Mode, and Copilot. You may find that you appear often in one and rarely in another. That gap is a clue. Bing powers Copilot and Googlebot powers Google’s AI features, so weak results in one engine can point to an indexing or listing problem with that search ecosystem.

Per-location breakdown

For local and regional businesses, location changes everything. A law firm in Lahore competes with different names than a law firm in Karachi. A restaurant group with branches in Austin and Dallas should measure each city separately.

Put the location in the question itself (“best family lawyer in Lahore”) so the answers are comparable. Then calculate share of voice per city or region. A brand can lead in one city and be absent in the next. Our guide to local AI visibility for “best service in city” questions covers this in detail.

Per-topic breakdown

Group questions by service or product line. A digital agency might be strong for “SEO agency” questions and invisible for “web design agency” questions. Topic-level numbers show you where to focus content and outreach.

How do you raise your AI share of voice?

No one can promise that an AI assistant will mention you. No AI company publishes its ranking formula. What you can do is improve the odds and then measure whether it worked.

1. Make sure AI search crawlers can reach your site. Blocking search crawlers such as OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot or Bingbot can keep you out of answers. Blocking training crawlers alone does not. See how to allow AI crawlers in robots.txt.
2. Study the competitor who leads. Look at the answers where they are named and you are not. What pages do the engines cite? Which directories, review sites or “best of” lists include them?
3. Get onto the sources AI engines cite. If an answer cites a list of top providers in your city and you are missing from it, that is a concrete gap. Our guide on which websites ChatGPT cites explains how to find these pages.
4. State your facts clearly and consistently. Use the same business name, address, phone and service descriptions on your website, Google Business Profile, Bing Places, Apple Business Connect and directories. This helps engines match everything to one business.
5. Add structured data that matches your visible content. Organization or LocalBusiness markup with sameAs links, address and areaServed helps machines understand who you are. See Organization and LocalBusiness schema for AI search.
6. Answer the questions people actually ask. Pages that directly answer a specific question are easier for engines to retrieve and quote. Clear FAQs help too.
7. Re-measure with the same question set. Changes take time to show up, and answers fluctuate. Compare trends across several checks before drawing conclusions.

What should an AI share of voice report include?

Whether it is for your own business or an agency client, a useful report is short and actionable:

• The question set used and when it was checked
• Mention rate and share of voice, overall and per engine
• A breakdown by location or topic where relevant
• The competitors who appear most often
• Gaps: questions where a competitor is named and you are not
• The sources cited in those answers
• Actions taken since the last report, and what changed

Avoid presenting a single percentage as the whole story. Context, trend and gaps matter more than the headline number.

How Onxeera helps

Onxeera’s Prompt Research tool lets you add the questions your customers ask, then asks ChatGPT, Perplexity, Gemini and Claude through their official APIs with web search on. It stores each answer, the brands mentioned, their order and the cited sources, so you have the raw data for share-of-voice calculations. Your mention rate appears only after at least 3 answers, gaps show where a competitor is named and you are not, and paid plans add auto-tracking so you can follow trends. To close those gaps, the Citation Builder shows which cited pages mention your competitors but not you. You can start on the free plan from the pricing page.

Frequently asked questions

Is AI share of voice the same as mention rate?

No. Mention rate is the share of answers that name you at all. Share of voice compares your mentions with all brand mentions across the same answers, including competitors. A business can have a decent mention rate but a low share of voice if several competitors are named more often.

How often should I measure share of voice in ChatGPT?

Measure on a regular schedule, such as monthly or weekly, using the same set of questions each time. AI answers fluctuate, so a single check is not reliable. Trends across several checks are more meaningful than any one result, especially after you make changes to your site or listings.

Can I measure competitor visibility in AI without paid tools?

Yes. You can ask each AI assistant your questions manually, record the brands named in a spreadsheet and apply the formula. It takes time and is harder to repeat consistently, which is why many teams use a tool that stores answers, brands and sources automatically.

Why does my share of voice differ between AI engines?

Each engine searches the web differently, relies on different sources and may favour different brands. Copilot relies on Bing, while Google's AI features rely on Google Search. A gap in one engine often points to missing listings, indexing issues or absent citations in that engine's ecosystem.