AI answers don't come with a rank you can look up. To know how your brand is doing, you ask the same questions again and again, in each market, and count what comes back. This guide explains the metrics GoGetCited uses, what each one tells you and what it doesn't, and how to read them without mistaking noise for news.
The building blocks
Every metric is counted over answers, so it helps to be precise about the terms.
- Prompt: a question customers ask an AI assistant, such as "best coffee subscription for espresso".
- Market: a country plus a language. The same prompt can run in several markets.
- Answer: one engine's full reply to one prompt in one market on one day.
Ten prompts checked daily in two markets on two engines give 40 answers a day, or 1,200 in 30 days. That pool is what the metrics below are calculated from.
Brand mentions are found by exact matching on each brand's names, aliases and domains. They are not guessed by AI, so the same answers always give the same counts. AI is used only where judgment is needed, such as rating sentiment and classifying source types. And only answers that exist are counted: when Google shows no AI Overview for a search, GoGetCited records that, but it doesn't count as an answer that left you out.
Brand metrics
Visibility
The share of answers that mention your brand. If 300 of 1,200 answers name you, your visibility is 25%.
Visibility is the first number to watch, because it answers the most basic question: are you in the answer at all? It also doesn't depend on which competitors you track.
Share of voice
Your brand's mentions divided by the mentions of all tracked brands. If the answers mention your tracked brands 1,000 times in total and 250 of those mentions are yours, your share of voice is 25%.
Share of voice is relative. Add a competitor that is named often and your share of voice falls, even if the answers haven't changed. Use it to compare yourself with a stable set of rivals, not as an absolute score.
Position
Your rank among the tracked brands, by order of first mention in the answer. If an answer names a competitor first and you second, your position in that answer is 2. GoGetCited averages it over the answers that mention you, so lower is better.
Position only counts answers where you appear. An average position of 1.5 at 5% visibility means you are named early in the few answers that mention you, not that you are winning. Always read position next to visibility.
Sentiment
The tone of each mention, on a scale from 0 to 100: 0 is very negative, 50 is neutral or purely factual, 100 is very positive. Each score comes with the reason. A low score with a reason like "mentions slow delivery" points to something specific you can check.
Sentiment is averaged over mentions, so a few strongly negative answers can hide inside a good average. Read the most negative snippets as well as the score.
Source metrics
Citation share and Used %
How often a domain or page is cited. In the Sources table, Used is the share of answers that cite a domain, and citation share is that domain's share of all citations. A site with a high Used % is one the answers rely on for your questions, whether or not it mentions you.
Own citation rate
The share of answers that cite your own site. It is separate from visibility. Your guide can be cited while the answer recommends someone else, and you can be named without your site being cited.
Gap
Answers where a competitor is named or cited and you aren't. Gap is the most practical number on this list: every gap answer points to a specific prompt, a specific competitor and the sources behind them. Those are places to publish, or places to get mentioned.
How to read the metrics together
No single metric tells the whole story. Some common combinations:
- Low visibility, good position. When you are named, you are named early, but you are missing from most answers. Look at the prompts where you are absent and the sources cited there.
- High visibility, poor position. You are in the list, but late. Compare what the first-named brands have that you don't: reviews, comparisons, specific facts.
- Rising share of voice, flat visibility. A competitor may have lost mentions while you gained none. Check visibility before you celebrate.
- High own citation rate, low visibility. Your content is used as a source, but the answer recommends others. Make your offer and its strengths clear on the pages that get cited.
- Gap concentrated in a few sources. A handful of review articles or threads drive most of the answers you miss. That is earned work: get onto those pages.
Then go one level down, from the number to the answers behind it. Reading ten real answers often explains a change faster than any chart.
Why answers vary from day to day
AI answers are not fixed. The same prompt can name different brands and cite different sources on consecutive days. The model's writing varies, the search results it reads change, and engines update their models. Answers also differ by market and language.
So treat each answer as one sample, and follow a few rules:
- Look at trends over 14 to 30 days. A single day tells you little.
- Make sure there are enough answers. If a metric rests on 20 answers, two answers changing moves it by 10 points. Base conclusions on enough prompts and markets.
- Compare periods of equal length. GoGetCited shows every metric's change against the previous period of equal length, so the last 30 days are compared with the 30 days before.
- Keep your prompt set stable. Adding or removing many prompts changes the metrics even if nothing else changed. Choosing prompts to track explains how to build a set that lasts.
- Read the answers behind any big move before you act on it.
Comparing engines and markets
Engines don't behave the same way. They search differently, cite different kinds of sources and name different numbers of brands. A brand can be named in most ChatGPT answers and rarely in AI Overviews for the same prompts. That isn't a contradiction. It reflects how each engine finds its sources.
The most reliable comparisons stay within one engine and one market: you against your competitors, or this month against last month. A matrix of brands by engine, and of brands by market, shows where you are strong and where a competitor owns the answer.
Markets need the same care. Each has its own language, its own competitors and its own sources. A brand that leads in one country can be missing in another, and the fix is often local: local review sites, pages in the local language, local facts.
A worked example
The numbers below are made up for illustration. Northwind Coffee, an example brand, tracks ten prompts about coffee subscriptions in the United States and the United Kingdom, on two engines. After 30 days there are 1,200 answers.
| Brand | Answers naming it | Visibility | Share of voice | Avg. position |
|---|---|---|---|---|
| Northwind Coffee | 300 | 25% | 28% | 2.3 |
| Copperline Roasters | 540 | 45% | 50% | 1.4 |
| Little Fern Coffee | 240 | 20% | 22% | 2.0 |
All three brands are made up too. Here is how Northwind's team might read the numbers:
- Visibility. Northwind is named in a quarter of the answers, Copperline in almost half. The first goal is to appear more often.
- Position. When Northwind is named, it usually comes after Copperline.
- Markets. Split by market, Northwind's visibility is 40% in the United Kingdom and 10% in the United States. The problem sits mostly in one market.
- Sources. Northwind's own site is cited in 96 answers, an own citation rate of 8%. One comparison article on a coffee review site is cited in 150 answers, and 90 of them name Copperline but not Northwind.
- Gap. 560 answers name or cite a competitor and not Northwind, and that one review article sits behind many of them.
The daily numbers moved a lot: one day Northwind's visibility was 35%, the next day 15%. With 40 answers a day, that is a difference of eight answers. Neither day says much on its own; the 30-day figure does.
So the team writes a US-focused page that answers "best coffee subscription for espresso" directly, contacts the review site with up-to-date facts, and compares the next 30 days with these.
How GoGetCited helps
GoGetCited calculates these metrics for your brand and every tracked competitor from each answer it records, by engine, market, topic and prompt, and shows each one's change against the previous period of equal length. Competitor benchmarking ranks the brands and lists the prompts they win that you don't. You can open any number to read the answers behind it, and the same figures go into reports and CSV exports.