Why doesn't "rank in ChatGPT" work as a metric?
In search engines, a ranking is fairly stable. You're third today and probably third tomorrow. AI answers don't behave that way. A study by SparkToro and Gumshoe in early 2026 ran the same 12 prompts nearly 3,000 times through ChatGPT, Claude and Google's AI. The chance of getting the same list of brands twice was under 1 in 100. The same list in the same order came up about 1 time in 1,000.
The same study found something more useful. While order was close to random, presence wasn't. Some brands appeared in 60 to 90 percent of answers for a given type of question. Its takeaway was blunt: repeated presence means something, exact rank doesn't.
Which AI visibility metrics should you track?
| Mention rate | Answers that name you ÷ all answers | How often buyers asking AI will hear your name at all |
|---|---|---|
| Share of voice | Your mentions ÷ all brand mentions in those answers | Your slice of the conversation compared with competitors |
| Average position | Where you appear in the list, when you appear | Useful as a rough signal only, since order changes a lot |
| Cited sources | Which websites the answers link to or draw from | Where to get mentioned to move the other numbers |
Mention rate and share of voice answer different questions. Mention rate is about reach: if 100 people ask, how many hear about you? Share of voice is about competition: when brands are named, how much of that attention is yours? A brand can have a high mention rate in a crowded category and still a modest share of voice.
A worked example
Say you sell project management software and track 20 buyer questions on three assistants. That's 60 answers per check. Across those answers, brands are named 180 times in total.
| Metric | Your brand | Biggest competitor |
|---|---|---|
| Answers that name the brand | 21 of 60 | 45 of 60 |
| Mention rate | 35% | 75% |
| Mentions out of 180 | 27 | 72 |
| Share of voice | 15% | 40% |
| Usually listed | 4th or 5th | 1st or 2nd |
The gap that matters here isn't that the competitor tends to be listed first. It's that they're in three answers out of four and you're in one out of three. That's the number to move, and the cited sources show where the competitor is getting its mentions from.
How to measure AI share of voice properly
- 1Write down the questions buyers really ask
Full questions, not keywords: "best CRM for a 10-person agency", "X vs Y for invoicing". Mix early research questions with comparison and pricing ones. 20 to 50 is a good start.
- 2Pick a fixed competitor set
Choose 3 to 8 brands and keep them the same between checks, or your share of voice will move for reasons that have nothing to do with you.
- 3Ask more than one assistant
ChatGPT, Gemini and Claude draw on different sources and often disagree. A brand can be strong in one and missing in another.
- 4Repeat on a schedule
Weekly or daily, at roughly the same time. Several runs of your most important questions give a steadier number than one run of many.
- 5Read the trend, not the day
A swing of a few points between two checks is often noise. Four weeks moving the same way is a signal.
What actually moves share of voice?
Share of voice follows the sources AI reads. When an assistant searches the web to answer a question, it builds its list from the pages it finds: roundups, review sites, forum threads, comparison articles and brand sites. Get onto more of those pages, with accurate descriptions of what you do, and your share tends to rise. Our guide on getting recommended by ChatGPT walks through the full process, and brand mentions vs backlinks explains why mentions carry so much weight.
Frequently asked questions
What is AI share of voice?
The percentage of brand mentions in AI-generated answers that go to your brand, measured across a fixed set of questions and a fixed group of competitors.
How is share of voice different from mention rate?
Mention rate is the share of answers that name you at all. Share of voice is your share of all brand mentions. One measures reach, the other measures your slice compared with competitors.
How many prompts do I need to measure AI visibility?
Start with 20 to 50 real buyer questions across at least two or three assistants, and repeat them regularly. More runs of your most important questions give steadier numbers.
Why do ChatGPT's recommendations change every time?
Answers are generated fresh each time and often draw on live search results, so the wording, the sources and the list of brands vary. That's why presence across many answers matters more than any single answer.
Can I track AI share of voice for free?
You can do it by hand in a spreadsheet, asking each assistant each question and counting mentions. Tools like GrowMarketingPro automate the asking, counting and trend charts.