Share of voice in AI answers, and how to compute it
Learn how to calculate AI share of voice, separate mentions from citations, and interpret a real result from 80 sampled answers.
By CiteGraph · 21 Sept 2026 · every number below has a source
A product named in 60% of answers has the strongest observed presence in this sample. Its own site can still appear in far fewer answers. That gap is where AI share of voice measurement starts.
The AI search visibility and AEO tools leaderboard is built from 80 answers to 10 buyer questions across 4 engines, according to CiteGraph's leaderboard for the week of 2026-08-24. Here is how to calculate the result, read its limits, and turn it into page-level work.
1. Set the answer count
Share of voice is the percentage of relevant answers that mention your product, brand or company. Use this formula:
Share of voice = answers naming your brand ÷ total relevant answers, expressed as a percentage
Set the buyer questions, engines and time period before you count anything. The leaderboard rows use the same question and engine mix, so the product comparison has a consistent basis.
Start with a fixed sample.
2. Keep mentions and citations in separate columns
An engine can name a product without linking to its website. It can also cite a page without making that product the main recommendation. Record those events separately, along with the question, engine, cited domain and cited page.
Calculate the share of answers naming each product. Then calculate own-site citation share separately: divide the number of answers that cite your own site by the total relevant answers, using the same denominator as above. The leaderboard says Otterly is named in 60% of answers and its own site is cited in 26%. Peec AI is named in 53% and its own site is cited in 18%, while Profound is named in 48% and its own site is cited in 20%.
Keep the two measures visible.
3. Work through a real share of voice example
Take Otterly from the leaderboard. The sample contains 80 answers, and Otterly is named in 60% of them.
Line by line:
Total answers: 80.
Share naming Otterly: 60%.
Convert the percentage to a decimal: 60% = 0.60.
Apply the decimal to the sample: 0.60 × 80 = 48 answers.
State the result: Otterly appears in 48 of the 80 sampled answers, giving it a 60% share of voice.
The leaderboard also says Otterly's own site is cited in 26%, with its pricing page cited most often. The calculation describes this set of questions, engines and runs.
The 48 answers sit inside Otterly's reported interval of 49 to 70 for the percentage estimate, so the observed count should not be treated as a permanent market position.
Share of sampled answers naming each AI SEO tool in the 80 answer leaderboard.
The figure shows observed answer presence across the products. Otterly leads at 60%, followed by Peec AI at 53% and Profound at 48%. AirOps and Rankscale are each at 9%. This is a ranking of mentions in the sample, not a product quality score.
Read the margin around 80 answers
The leaderboard gives Otterly an interval of 49 to 70, Peec AI an interval of 42 to 63, and Profound an interval of 37 to 58. These ranges show how much the estimate could vary around the observed share.
A reasonable rule of thumb is to treat overlapping intervals as a tie for decision-making rather than claiming a clear lead. Collect more comparable answers before making a stronger comparison.
This is the limit of the data: we measure samples, we print intervals, and we do not know what we have not measured.
Our wider research records 64,667 citations across 9,791 AI answers, from 77 scans of 44 sites in 48 categories. Those scans asked 507 distinct buyer questions, and the engines cited 9,192 distinct domains and 21,465 distinct pages. The leaderboard ranks products in one category; this wider dataset tracks citation patterns across categories to spot broader trends.
Use citation share as a second measure
Share of voice tells you how often an answer names you. Citation share tells you how often the answer uses your site or page as support. Keep both measures because a product can be widely mentioned while its own pages appear less often as sources.
Otterly's 60% mention share and 26% own-site citation share are different numbers for a reason: being named is not the same as being the source. Inspect the cited URLs and the wording around each mention.
Measure the two surfaces side by side.
Review the page evidence
The Otterly pricing page is the most cited Otterly page in the leaderboard. The ledger records no page-content versions or earlier citation status, so it cannot support a before-and-after description of that page. Do not claim that the page changed based on this dataset.
To create that comparison, save the page version and cited-answer records before making a change, then compare them with a later run. Add the question, engine, mention status, citation status and cited URL to the record. Our guide to measuring AI citations and methodology for the measurements explain that process.
The useful change is evidence you can inspect.
Keep the raw answers behind the percentage
A headline share can hide differences between buyer questions and engines. The leaderboard covers 10 questions across 4 engines, so a result may reflect that particular mix. A weekly sample can also catch a temporary source change or miss a question that matters to your sales team.
Save the exact question, engine, date, mention status, citation status and cited URL. When the percentage moves, you can check whether the cause was a new mention, a lost citation, a changed question mix or a different source page. The wider weekly research gives context, while the raw answers explain the individual result.
Ask next: which buyer question and which page should change next? The answer lives in the raw AI responses and cited URLs behind the percentage.
Questions people ask
What's the formula for AI share of voice, and what do I need before I can use it?+
Divide the relevant answers that name your brand by the total relevant answers, then express the result as a percentage. Set the questions, engines and time period first, and keep citation counts separate.
What does a 60% AI share of voice mean?+
It means the brand appeared in 60% of the sampled answers to the chosen questions. In the leaderboard sample of 80 answers, that works out to 48 answers, and the result applies to that sample.
How many answers do I need before I can trust the comparison?+
There is no universal answer count that makes a result final. Use the interval around the observed share, and as a reasonable rule of thumb, treat overlapping intervals as a tie before collecting more comparable answers.
Is AI share of voice the same as AI citation share?+
No. Share of voice counts answers that name you, while citation share counts answers that use your site or page as a source. A product can have a high mention share and a lower own-site citation share.
Which pages should I review after measuring AI share of voice?+
Start with the pages engines cite most often for your brand and competing brands. Compare those URLs with the buyer questions where you are mentioned or cited, then save the raw answers so you can check what changed later.