How to measure share of voice in AI chatbot answers

In classic marketing, share of voice is how much of the conversation in your market is about you rather than your competitors. In AI search the conversation happens inside ChatGPT, Gemini and Claude answers. So AI share of voice asks: when buyers ask these assistants about your category, how often are you the one recommended?

Here's how to measure it properly.

What to count

The unit is an answer to a buyer question. Not a keyword, not a mention on a website. For each answer, you note which products are named and in what order.

You need three things:

1. A fixed set of buyer questions. 10 to 50 questions buyers ask just before choosing, like "best [category] for [customer]" or "[leader] alternatives". Keep the wording the same over time. See how to choose buyer questions.
2. A set of assistants. Usually ChatGPT, Gemini and Claude. Each question asked on each assistant is one "cell".
3. Your brand and your rivals. The products you want to compare.

The simple formula

The simplest version counts mentions:

Share of voice = answers that name you ÷ total answers

If you ask 10 questions on 2 assistants and you're named in 8 of the 20 answers, your share of voice is 40%.

To compare with rivals, do the same for each one. If Rival A is named in 16 of 20, they have 80%. Note these don't add up to 100%, because one answer can name several products.

Why position matters

Being named first and being named fifth are very different. Buyers read from the top and often go with one of the first suggestions. So a better measure weights each mention by position.

Citemark's visibility score uses these weights:

Average the weights across all answers and multiply by 100. Being named first in every answer gives 100; never being named gives 0. How the AI visibility score works explains the details, including why failed answers are left out instead of counted as zero.

Share of voice against rivals

The score above is just for you. For share of voice you calculate the same score for each rival, from the same answers, and put them side by side:

This shows who owns the conversation and by how much. It also shows close races, like You and Rival B above, where a small push could move you ahead.

Getting a number you can trust

AI answers vary from run to run, so small samples are noisy. To get a steady number:

• Ask many times. More answers means less noise. Daily checks on your key questions add up quickly.
• Use a 7-day average. A rolling average smooths out daily swings so you can see the real trend.
• Don't change the questions often. Changing them changes the measure. Add new ones alongside old ones instead.
• Look at each assistant separately too. A rise on ChatGPT can hide a fall on Gemini.

Why AI gives different answers each time covers the variation and how to handle it.

Breaking it down

The overall number is useful for reporting. The breakdowns are useful for deciding what to do:

• By question: which questions you win, lose, or that nobody wins yet.
• By assistant: where you're strong and where you're missing.
• By rival: who you lose to most often.
• By cited source: which pages are behind the answers you lose.

Reporting it to your team

Share of voice is easy to explain, which makes it a good number for a weekly update or a board slide. A simple format that works:

1. The headline: "We're named in 45% of AI answers to our top buyer questions, up from 38% last month. Rival A is at 70%."
2. One win: a question you now win that you didn't before, and what you did to win it.
3. One gap: the question where you lose most, the rival who wins it, and the page that's cited.
4. Next step: the one thing you'll do about that gap this month.

Keep the questions and the method the same from month to month so the numbers stay comparable. If you change the question list, say so, and show the old and new numbers side by side for one month.

What a good share of voice looks like

There's no universal target, because it depends on how crowded your category is. A useful rule of thumb: aim to be named in at least half the answers to your most important questions, and to be in the top three when you are named. If a single rival is named in nearly every answer, closing that gap is usually worth more than adding new questions.

Measuring it with Citemark

Citemark measures this for you. It asks your buyer questions on ChatGPT and Gemini (and Claude on Pro), with your most important questions checked every day, and keeps every answer. The dashboard shows your weighted score and trend, a brand table comparing you with each rival, and the sources behind each answer.

Start measuring

Run the free AI visibility check to see your share of answers for 5 buyer questions in about a minute. To track share of voice against your rivals daily, start a 7-day Pro trial with no card.