Blog Guide
AI visibility metrics: visibility share, citation share and position
What AI visibility metrics mean and how to calculate them: visibility share, citation share, average position, sentiment, competitor shares and cited sources.
AI visibility is measured with a few shares taken over a sample of AI answers. Visibility share is the share of answers that name your brand. Citation share is the share that link to your site. Average position is where you appear in the list of brands named, counted only in answers that name you. Calculate the same numbers for your competitors and read them over several weeks; a single week is a sample, not a verdict.
This is the reference for what each number means, how to calculate it and how to read it. For the hands-on routine of choosing questions, asking them and logging the answers, see how to track brand mentions in ChatGPT.
Why AI visibility is a share, not a rank
In classic search, a page holds a fairly stable position for a keyword. AI answers do not: the same question asked twice can name different brands in a different order, and location, language, a signed-in account and small changes of wording all change the result. No engine publishes a rank either. So every AI visibility metric is a share or an average across many answers: the questions your buyers ask, on each engine you care about, collected on a schedule.
The core metrics
Visibility share
Answers that mention your brand ÷ all answers analysed
If you track 25 questions on 2 engines once a week, that is 50 answers. If 12 of them mention you, your visibility share is 24%. This is the headline number: how often you make the shortlist at all. Count a mention when the answer names your brand or a name buyers use for it, such as a product name or an old brand name.
Citation share
Answers that cite a page on your domain ÷ all answers analysed
A mention is your brand named in the answer. A citation is a link to your site as one of the answer's sources. An answer can mention you without linking to you, and link to you without naming you in the text, so the two shares move separately. Citation share tells you whether the engine treats your own pages as a source. Count any subdomain, such as a blog or help centre.
Average position
The mean of your position across the answers that mention you
Position is your place in the list of brands an answer names: position 1 is the first brand named. Average only over answers where you appear; absences are already counted in visibility share, and giving them a position would pull the average around in a way that means nothing.
Read the two numbers together. A brand with a visibility share of 20% and an average position of 1.5 is rarely shortlisted but leads when it is. A brand at 60% and 4.0 is usually shortlisted but rarely first. They need different fixes.
Sentiment
For each mention, note the tone: positive, neutral, mixed or negative. Visibility share counts every mention the same, so a brand can be named often and described poorly. Count the answers that describe you negatively or with mixed feelings, and read them: they usually repeat a claim from a page you can find and answer.
Competitor shares
Run the same calculations for three to five competitors, on the same answers. A 24% visibility share is hard to judge on its own. Next to a competitor at 60% on the same questions, it is easy to read.
Most-cited sources
Rank every domain cited across the week's answers by how often it appears, and mark the ones that cited your site. The list usually mixes review and directory sites, community threads, publishers and competitors' own pages. These are the pages the engines learn from, and pages you can act on.
Break the numbers down
A single overall share hides where you win and lose. Split it two ways:
- By engine. ChatGPT, Gemini, Perplexity, Google's AI Overviews and AI Mode draw on different sources and often disagree. A brand can be strong on one and missing from another.
- By question. A question that names you on no engine, in any sample, is the clearest gap you have: nothing on the web answers it with you in it.
Keep the numbers honest
- Exclude failures. If an answer did not come back or could not be read, leave it out of the totals instead of counting it as "not mentioned".
- Flag thin weeks. If far fewer answers than usual were analysed, mark the week as partial and give it less weight.
- Mind the sample size. In a sample of 50 answers, each answer is 2 percentage points, so a move of a few points can be one or two answers.
- Read the trend, not the week. A move from 24% to 27% in one week may be noise. A move from 24% to 35% over six weeks, after you published three comparison pages, is a signal.
Reading the numbers together
| What you see | What it usually means | Where to look |
|---|---|---|
| Low visibility, low citation | The engines know little about you for these questions | Your pages for the lost questions, and the most-cited sources that leave you out |
| High visibility, low citation | The engines name you but learn about you from other sites | The cited sources: are they accurate about you? |
| Visibility rising, position falling | You are on more shortlists, but lower down | Comparison and alternatives pages that rank you against the leaders |
| Strong on one engine, weak on another | The engines draw on different sources | The cited sources per engine |
| Frequent mentions, poor sentiment | You are known, but described poorly | The negative and mixed answers, and the pages they cite |
How CiteGaps calculates them
CiteGaps applies these definitions every week, on every engine in your plan, per engine and per question, with the change since the week before. A few details:
- Mentions match your brand name or any alias you add; citations count any subdomain of your site.
- A competitor's citations count on its domain when you give one; otherwise a domain counts when its main label matches the name, as hubspot.com does for HubSpot.
- Top sources are the 25 domains cited most often in the week.
- Three or more negative or mixed answers in a week raise a "weak sentiment" recommendation.
- A week in which fewer than 60% of the expected answers could be analysed is marked partial.
How CiteGaps works covers the weekly run and the recommendations, and the free check tries one question now.
Questions
Is visibility share the same as share of voice?
Not quite. Visibility share divides the answers that name you by all answers analysed, so several brands can score high on the same questions. Share of voice usually divides your mentions by all brand mentions, so the shares add up to 100%. Visibility share answers "how often am I on the shortlist?"; share of voice answers "how much of the conversation is mine?".
What is a good visibility share?
There is no universal benchmark: it depends on your category, the questions you track and how many brands compete for them. Compare yourself with your competitors on the same questions, and watch your own trend over several weeks.