gptagency.io

Methodology

How would we know? Like this.

The Visibility Score measures how present a business is in AI answers: a fixed panel of hundreds of real customer questions is put to four AI assistants every month, each question multiple times. What counts is whether the business is mentioned, recommended, or recommended first. The result is a number from 0 to 100 – measured with the same yardstick month after month, and therefore honestly comparable.

The measurement in four steps

Fragen-Panel 120 echte Kundenfragen 4 KI-Assistenten jede Frage 3× ChatGPT · 3 Runs Claude · 3 Runs Gemini · 3 Runs Perplexity · 3 Runs Auswertung genannt? wo? wie? Score 0–100, jeden Monat 40

1 / 4

The question panel

Hundreds of questions the way customers really ask them – from the problem question ("Where can I find…") to the comparison question to the question about your name. The panel is built once and then stays fixed: only if you ask the same things every month may you compare the months.

2 / 4

Four AIs, three runs

Every question goes to ChatGPT, Claude, Gemini, and Perplexity – three times each. AI answers vary; a single measurement would be a coin flip. Only the average across several runs is a reliable value.

3 / 4

The evaluation

Every answer is read and marked: do you appear? Are you actively recommended? Are you first? And who else is named? Every answer is stored verbatim – you can re-read every single verdict.

4 / 4

The score

All verdicts combine into a number from 0 to 100. Mentioned counts, recommended counts more, first counts most. Added to that: the trend across months, your share of all recommendations, and the breakdown by question type.

The weighting

Mentioned. Recommended. First.

Mentioned

Your name appears in the answer. The base level – you exist for the AI.

Recommended

The AI actively advises choosing you, with a reason. This is the level that brings customers.

First

You are the first recommendation in the answer. The position with the greatest weight in the score.

Honesty

What the score can do – and what it cannot

  • Make months comparable: same panel, same AIs, same yardstick
  • Show gaps precisely: every lost question is a concrete work order
  • Prove impact: which measure won which questions
  • Put competition in context: who gets recommended instead, and where
  • Predict revenue – it measures visibility, not bookings
  • Guarantee any single AI answer – AIs remain probability machines, which is why we measure repeatedly
  • Rate other channels – Google rankings and ads have their own yardsticks

Want to see this for your business?

The visibility check is the small edition of this measurement – free, results in 48 hours.