GEO toolAI share of voice calculator: how often AI names you.
There are no rankings inside an AI answer, only who gets named and who gets cited. Run a fixed set of prompts in each assistant, log the brands and sources, and paste the results here. You get share of voice against competitors, by engine and run by run, with the margin of error that tells you whether a change is real. Your data stays in your browser.
Free · no signup · updated 16 October 2026
Paste your prompt panel results
One row per answer: the prompt, the engine, the brands it named in order (separated by |), the URLs it cited, and the run date. A CSV or a copy from Google Sheets both work. Nothing is uploaded.
Share of voice by brand
By engine
| Engine | Answers | Names you (± margin) | Share of voice | Weighted share | Cites your site |
|---|
Run by run
| Run | Answers | Names you | Share of voice | Weighted share | Cites your site |
|---|
Most-cited sites
| Site | Answers citing it | Citation share |
|---|
How do you calculate AI share of voice?
Run a fixed set of category prompts in each AI assistant and record which brands each answer names. Your share of voice is your mentions divided by the mentions of all tracked brands. Mention rate is the share of answers that name you, and citation share is the share of answers that link to your site.
The method, in the open.
Each row is one answer. Columns are matched by name (prompt, engine, brands, cited, date), or read in that order if there is no header. Brands are split on |, ; or commas inside the cell; a brand counts once per answer.
- Share of voice = your mentions ÷ mentions of every brand in the data.
- Mention rate = answers that name you ÷ all answers, with a 95% margin of error (a Wilson interval, which stays sensible at 0% and 100%).
- Position-weighted share: the first brand named scores 3, the second 2, the third 1; your points ÷ all points.
- Citation share = answers citing your domain (or a subdomain) ÷ all answers.
With several runs, the headline uses the latest one and the run-by-run table shows the trend. A change smaller than the margin of error is flagged as noise.
What the numbers look like.
Using the sample results. Four prompts were run in ChatGPT, Perplexity and Gemini in September and October: 24 answers. In October Acme Growth is named in 9 of 12 answers (75%), up from 4 of 12, and holds a 36% share of voice against Northstar's 32%. Its site is cited in 6 of 12 answers. The jump is bigger than the ±22-point margin for 12 answers, so it is probably real. But Gemini names Acme in only 1 of 4 answers and never cites it, so that is the engine to work on next.
What this is based on
- Asked the same prompt repeatedly, ChatGPT, Claude and Google's AI gave the same brand list less than 1% of the time; how often a brand appears across many runs was far steadier than its position. SparkToro and Gumshoe, via Search Engine Land
- Branded web mentions correlated 0.664 with ChatGPT visibility across 75,000 brands, more than backlinks or URL rating. Ahrefs
What it does not tell you
- The results are only as good as the prompt panel. Keep the same prompts each month, and report category, comparison and brand prompts separately.
- Brand names must be written the same way in every row; "Acme" and "Acme Growth" count as two brands. Clean them in the sheet first.
- Answers vary between runs, so small panels give wide margins. Aim for 50+ prompts per engine and rerun key prompts.
- Position weighting assumes the order brands were named reflects prominence, which isn't always true for tables or grouped lists.
Questions people ask.
What is AI share of voice?
The share of brand mentions in AI answers, across a fixed set of prompts, that belong to your brand rather than competitors.
What is the difference between share of voice and mention rate?
Mention rate is how often you are named at all. Share of voice is your slice of all brand mentions, so it also moves when competitors gain or lose.
How many prompts do I need?
Enough for the margin of error to be smaller than the changes you care about. With 50 answers per engine it is about ±13 points at worst; with 200, about ±7.
How should I format my data?
One row per answer with columns date, engine, prompt, brands (in the order named, separated by |) and cited (URLs or domains separated by |). Our 50-prompt tracker article has a template.
Is my data uploaded?
No. The file is read and calculated in your browser. Nothing is sent to our servers.
Keep checking.
AI citability checker
Paste a page and see how easy it is for ChatGPT, Perplexity and AI Overviews to quote: a direct answer, statistics, sources, quotes, dates and prompt fit. Every check links to its evidence.
Entity consistency checker
Paste your homepage, about page and app listing and see where they disagree about your name, what you do, profiles, contacts, founding year and location. Runs in your browser.
GA4 AI traffic regex generator
Track ChatGPT, Perplexity, Claude, Gemini and Copilot visits in GA4. Generates the regex, channel group steps, a Looker Studio formula and a GTM variable.
Get named in the answers your buyers read.
We build your prompt panel, track it across assistants every month, and fix the pages and mentions behind the gaps.