GEO consoleGEO consoleA metric is only worth something if you understand how it's produced. Here, with no black box, is how we generate the questions, query the assistants and compute your share of voice — and where the limits are. Last updated: June 11, 2026.
We start from your site. GEO console analyzes your business, your offers, your audience and your competitors, then turns that understanding into real customer questions, phrased in natural language — the way people talk to an assistant. We don't test keywords, but the real requests where your brand should appear.
The goal is relevance: an off-topic question tells you nothing useful. We cover the intents where a recommendation is decided — direct recommendation requests, comparisons, alternatives to a competitor, local or niche needs.
Each question is asked to the five engines — ChatGPT, Gemini, Claude, Perplexity and Grok. And not just once: we repeat every prompt over several runs to smooth out the models' natural variance. This gives us a citation frequency, more reliable than a single attempt left to chance.
Each assistant forms its own idea of your brand: visibility is won engine by engine, and we track them side by side. The raw answers are kept, so every number stays traceable back to the source that produced it.
We don't publish a fixed number of runs: it depends on the engine and the context. What matters is the principle — repeat enough to tell a stable signal apart from mere randomness.
For each generated answer we extract four signals: whether your brand is cited, at what position it appears, your share of voice against the competitors cited, and the trigger prompts — the questions that make you appear, or not. Knowing you're cited isn't enough: you need to know where and next to whom.
Share of voice measures your relative weight against competitors across the whole panel of prompts. We consolidate your citations and those of other brands, across every engine and every run, into a proportion: how much space do you occupy in the answers, compared with those fighting for the same recommendations?
Concretely, a brand cited often, early in the answer and across many questions weighs more than one mentioned rarely and at the end of the list. Share of voice sums up that balance into a clear, search-console-style metric, available overall and assistant by assistant.
It's a relative metric, specific to your panel of questions: it reflects your standing against competitors on the requests tested, at a given moment. It isn't a universal web ranking, and we present it as such — a benchmark to track your progress, not an absolute truth.
No AI visibility metric is perfect, and we'd rather say so. Engines vary from one run to the next, coverage is never guaranteed, and each measurement reflects a given time window. Our numbers are reliable, reproducible benchmarks — not fixed certainties nor promises of results.
Models are probabilistic: the same question can yield different answers. Multi-runs reduce this noise but don't remove it — slight variations between two measurements are expected, and don't necessarily signal a real change.
We measure your presence, we don't guarantee it. Being cited depends on your content, your reputation and the sources the models rely on. GEO console shows you where you're missing and what to improve, without promising full coverage.
Each measurement is a snapshot at a given moment. Models evolve, the web changes, your competitors move. That's why AI visibility is tracked over time: a trend across several measurements beats an isolated snapshot.
Clear, search-console-style indicators to decide without jargon or a black box.
How often your brand is cited across all questions and runs, engine by engine.
Where you appear in the answer: being named first doesn't carry the same weight as a mention at the end of the list.
Your relative weight against the competitors cited on the same panel of prompts, overall and per assistant.
The questions that cite you — and those that cite your competitors in your place, so you know where to act.
What people ask us most often about how we measure.
Not perfectly, and that's normal: AI assistants are probabilistic, their answers vary from one run to the next. That's why we repeat each question over several runs to smooth out that variance and measure a citation frequency rather than a one-off result. Slight changes between two measurements are expected.
No. GEO console is a measurement and diagnostic tool, not a guarantee of presence. We show where you're cited, where you aren't, and what to improve. AI assistant coverage cannot be guaranteed, as it depends on your content, your reputation and the sources the models rely on.
Because a single answer isn't reliable. A model can cite you once and forget you the next time. By repeating each prompt over several runs, we turn a fragile yes/no into a stable citation rate, more representative of your real AI visibility across the panel of questions tested.
Enter your domain and get a first preview of your AI visibility — free, no credit card.
Test my site