How-to guide
How to measure your AI visibility, step by step
To measure your AI visibility, ask several assistants the buying questions in your category, repeat each one several times and count: how many answers include you (mention rate), what share of mentions is yours (share of voice), how many cite you as a source, in what position and with what tone. GeoRush automates this method across 9 engines.
Updated on
Why measuring AI is different
In SEO you can check your ranking for a keyword and get a fairly stable number. In AI you cannot: the same question can produce different answers on two consecutive attempts, each assistant recommends in its own way, and answers can change by country, language and, in personal accounts, the user’s history.
That makes AI visibility measurement a statistical exercise: a representative set of questions, repeated across several engines, from which you calculate frequencies with their margin of error. A screenshot of one conversation is an anecdote, not data.
The method in 6 steps
- 1
Define your question set
20 to 50 real buying questions grouped by topic: “best X for Y”, “alternatives to Z”, “X vs Y”, “how much does X cost”, “X in [city]”. Include questions that do not name your brand: what matters is whether you appear when nobody mentions you.
- 2
Choose engines, country and language
Measure the assistants your customers use (ChatGPT, Gemini, Perplexity, AI Overviews…) in the market and language where you sell. Answers for the US and the UK can differ.
- 3
Repeat each question
Ask each question several times per engine (for example 3 to 5) in a neutral setting: logged out or via API, so personal history does not interfere.
- 4
Record every answer
For each one: which brands appear and in what order, whether yours does, the tone, which sources are cited and whether any is your site. Keep the full text.
- 5
Calculate the metrics
Mention rate, share of voice, citation rate, average position and sentiment, per engine and per topic. Add the margin of error.
- 6
Repeat at fixed intervals
Weekly or monthly, with the same questions, to see trends and the effect of your changes.
The metrics and how to calculate them
| Metric | Calculation | What it tells you |
|---|---|---|
| Mention rate | Answers that name you ÷ total answers. | How often you exist for AI in your category. |
| Share of voice | Your mentions ÷ mentions of all brands. | Your weight relative to competitors in the same answers. |
| Citation rate | Answers that link your domain ÷ total answers. | Whether AI uses your site as a source of information. |
| Average position | Average place you hold when named. | Whether you are recommended first or at the end of the list. |
| Sentiment | Share of positive, neutral and negative mentions. | How you are described, not just whether you are named. |
| Cited sources | Domains that appear as sources, by type. | Where you need a presence in order to be mentioned. |
The margin of error matters (a lot)
With few answers, a percentage misleads. If you appear in 6 out of 20 answers, your mention rate is 30%, but the plausible range (95% confidence interval, Wilson method) runs from roughly 15% to 52%. With 60 out of 200, the rate is still 30%, but the range narrows to roughly 24%–37%.
The practical consequence: if you go from 30% to 40% on 20 answers, you cannot claim you improved; on 200, you probably can. Before celebrating or worrying about a change, check whether the ranges overlap. GeoRush shows mention rate with this interval so you do not make decisions based on noise.
Three ways to do it
| Option | Pros | Limitations |
|---|---|---|
| By hand in the chat | Free, immediate, good for a first impression. | Few answers, shaped by your history, hard to repeat the same way every week. |
| Spreadsheet + API | Full control and repeatable. | Requires coding, paying for each engine’s API, analyzing every answer’s text and maintaining it all. |
| Specialized tool | Automates questions, repeats, analysis and history across engines. | Costs money; check how it obtains answers and which engines it covers. |
Complementary signals
Answers are the primary measure, but other signals are worth cross-checking:
- AI referral traffic: visits from chatgpt.com, perplexity.ai, gemini.google.com and similar. ChatGPT adds
utm_source=chatgpt.comto its links. - AI crawler hits in your server logs: if OAI-SearchBot or PerplexityBot never visit, they are unlikely to cite you.
- Search Console: includes AI Overviews and AI Mode traffic within web search, without separating it; useful for cross-checking with your rankings.
- Accuracy of what is said: compare AI claims about your brand (prices, features, location) with your official facts.
How GeoRush does it
GeoRush applies this method automatically across 9 engines: ChatGPT, Gemini, Claude, Perplexity, Grok, DeepSeek, Le Chat, Google AI Overviews and Google AI Mode. Chat engines are queried via API with their official models and web search enabled; AI Overviews and AI Mode are read from real Google results. Each question is repeated several times, and Claude (Anthropic) analyzes every answer: brands and order, tone, themes, and claims about your brand checked against your official facts.
From that it calculates mention rate with its Wilson interval, share of voice, citation rate, average position and sentiment, and rolls them into the GEO Score (0–100). You can start with prompt research to build your question set, follow it in prompt tracking and share results with reports. Traffic from assistants and AI crawler hits live in web analytics. Plans and credits are on pricing.
Start measuring today
The free check asks one AI engine three buying questions and audits your site. No account, about a minute.
Sources
Official documentation and primary sources checked on November 16, 2026. These policies change often: check the source before making important decisions.
Frequently asked questions
Keep reading
What is GEO
The complete guide: what GEO is, how AI chooses what to recommend, how to measure it and how to improve.
Appear in ChatGPT
Step by step to get ChatGPT to mention your brand and cite your site, and how to check it.
GEO glossary
34 GEO terms clearly defined: concepts, metrics, crawlers and technical data.
AI visibility
Mention rate with a 95% range, share of voice, citations and GEO Score across 9 AI engines.