Features · AI visibility

AI visibility: how often AI assistants name you, and why

GeoRush measures your AI visibility by asking 9 assistants (ChatGPT, Gemini, Claude, Perplexity, Grok, DeepSeek, Le Chat, Google AI Overviews and Google AI Mode) the questions your buyers ask. It shows the share of answers that mention you, your position, whether your site is cited, and rolls it into a 0–100 GEO Score.

Updated on

The problem it solves

More and more buyers ask an AI assistant for a shortlist before they ever open a search results page. If ChatGPT recommends three vendors and you are not one of them, nothing in a classic SEO tool tells you: there is no ranking to check and no lost click to see.

GeoRush turns that blind spot into numbers. This is the foundation of GEO (generative engine optimization): knowing, with data, whether AI recommends you, against whom, on which engine and based on which sources. You use those numbers to decide where to work, and later to prove whether the work paid off.

How it works

  1. 1

    Pick your questions and competitors

    Add the questions your customers actually ask, such as “What's the best payroll software for a 20-person company in the US?”, and the competitors you want to benchmark. Start from GeoRush's suggestions or from Prompt research.

  2. 2

    GeoRush asks every engine

    Chat engines are queried through their official models via API with web search turned on, so answers cite sources. Google AI Overviews and AI Mode are read from real Google results. Every answer records how it was obtained.

  3. 3

    Each question is asked several times

    AI answers vary. By default every question runs 3 times per engine in each sweep (configurable from 1 to 5), and sweeps repeat 1 to 7 times a week. You see how often you really show up, not a one-off snapshot.

  4. 4

    Claude reads every answer

    Claude (Anthropic) extracts which brands appear and in what order, the tone, the themes linked to each brand, the sources cited, and new brands the AI recommends that you don't track yet.

  5. 5

    Metrics and GEO Score are calculated

    Using every successful answer in the period (7, 30 or 90 days), GeoRush computes mention rate with its range, share of voice, citation rate, average position and sentiment, and combines them into the GEO Score.

Every metric, defined

MetricWhat it tells youHow it is calculated
Mention rateHow often AI names you.Answers that mention you ÷ answers collected, with a 95% confidence range (Wilson interval).
Share of voiceYour slice of the conversation versus other brands.Your mentions ÷ all brand mentions in the answers.
Citation rateWhether AI uses your website as a source.Answers that link to your site ÷ answers collected.
Average positionWhere you appear when you are named.Average order of appearance (1 = first brand named), plus the share of times you come first.
SentimentHow AI talks about you.Positive, neutral and negative mentions, as classified by Claude.
GEO ScoreA 0–100 summary of where you stand.Weighted average of 5 pillars (next table). It is an estimated score designed for tracking progress.

Rates only use answers that were collected successfully. Failed answers don't count and don't use credits.

What goes into the GEO Score

The GEO Score combines five pillars. Each one has its own 0–100 score, so you can see where you are strong and where the biggest gain is.

PillarWeightWhat it measures
Visibility30%How many answers name you, compared with a 60% benchmark and with the category leader.
Citation20%Answers that link to your site. Without citations, AI describes you using third-party data.
Technical readiness20%AI crawlers allowed, llms.txt, structured data and pages that are easy to cite.
Share and position15%Your share of all mentions and how often you are named first.
Reputation15%Tone of mentions and whether what AI says about you is accurate.

GeoRush also gives a letter grade: A (80+), B (65–79), C (50–64), D (35–49) and E (below 35).

Why questions are asked more than once

Ask ChatGPT the same thing twice and you may get two different shortlists. A single query proves very little. GeoRush repeats every question on every engine and reports every percentage with a 95% confidence range.

For example, if you are named in 12 of 30 answers, the rate is 40% but the likely range is roughly 25% to 58%. With 48 of 120 answers the rate is still 40% and the range narrows to about 32–49%. Below 30 answers, GeoRush flags the figure as low confidence. It only reports that your visibility went up or down when the ranges for the two periods don't overlap, so noise is never sold as a trend.

What you see in GeoRush

  • GEO Score and grade, with a weekly trend line and a one-line summary of your strongest pillar and your biggest opportunity.
  • Mention rate per engine, so you know which assistant recommends you and which one barely does.
  • Weekly summary and insights: statistically clear gains or drops, rivals pulling ahead, topics another brand dominates, false claims about you.
  • What changed compared with the previous period.
  • Every original answer, word for word, in Prompt tracking.
  • The tasks that follow from the data in the Action plan.

Does AI name you when buyers ask about your category?

Run the free check: we ask one AI engine three buying questions about your category and give you an A–E grade. No account needed.

Frequently asked questions

Nine: ChatGPT, Gemini, Claude, Perplexity, Grok, DeepSeek, Le Chat (Mistral), Google AI Overviews and Google AI Mode. Every plan, including Free, includes all engines.

Mention rate is the percentage of AI answers that name your brand for the questions you track. GeoRush always shows it with a 95% confidence range, calculated with the Wilson interval.

No. It is an estimated 0–100 score that combines five measured pillars: visibility, share and position, citation, reputation and technical readiness. Use it to track progress and to decide which pillar to improve first.

Because AI assistants don't answer the same way every time. Repeating each question reveals how often you are really mentioned and makes a confidence range possible.

They are real. Chat engines are queried through their official models via API with web search on; Google AI Overviews and AI Mode are read from real Google results. Each answer shows how it was obtained.

It depends on the engine: 1 credit per answer for ChatGPT, Perplexity and DeepSeek; 2 for Gemini, Grok, Le Chat, Google AI Overviews and AI Mode; 6 for Claude. Claude's analysis is included. See pricing.

Keep reading