How it works

1. 01The question set is built
   Questions are prepared from your sector and what you actually sell — the kind a real customer would ask. Your brand name does NOT appear in them. Name yourself and the assistant will talk about you; that is an echo, not a measurement.
2. 02The questions go to the assistants
   The same question set is put to ChatGPT, Claude, Gemini and Perplexity separately. Every answer each assistant gives is stored in full.
3. 03Answers become structured measurements
   The same signals are extracted from every answer: was the brand mentioned, was it recommended, where was it named, what was the tone. Free text becomes numbers here.
4. 04Competitors are extracted
   The brands named in your place are collected from the answers. You do not supply the competitor list — who the assistants recommend is already in the answers.
5. 05Misinformation is flagged
   Untrue statements an assistant makes about the brand are flagged separately. This is independent of visibility and usually more urgent.
6. 06Findings become a report
   Measured findings, site-audit findings and Gekopilot's own interpretation carry separate badges. Where evidence cannot be shown, no cause is asserted.
7. 07The measurement repeats
   Under annual monitoring the same question set runs again every month. A single measurement is a starting point; direction only shows through repetition.

• My competitor appears in AI answers — why doesn't my brand?

You are notified by email when the report is ready, and it can be downloaded as a PDF. The formula behind the measurement and its limits are written out on the methodology page.

• Full methodology
• What is AI visibility?