GEO — Generative Engine Optimization
Your buyers ask an assistant. Does it mention you?
Search rankings tell you nothing about what happens inside a chat window. This runs the questions your customers actually type past several AI models and records, one answer at a time, whether your company gets named.
You get a PDF listing every question you were missing from, the most likely reason for each, and what to publish to fix it.
What actually happens
Six stages. No step is hidden, because the method is the part worth trusting.
-
01
Read the site
A few pages, prioritising what you do, who you serve and what it costs. Nothing is stored.
-
02
Write buyer questions
Five researchers work in parallel, each covering a different stage of the buying journey, in Arabic and English.
-
03
Cut the duplicates
Independent researchers converge. Near-identical questions are merged before anything expensive happens.
-
04
Judge relevance
A judge scores every question on whether it would be reasonable for you to appear, and keeps the strongest, balanced across both languages.
-
05
Ask the assistants
Each surviving question goes to every model on the panel, independently, as a real user would type it.
-
06
Diagnose and recommend
For each answer you were missing from, work out why and who was named instead — then turn the pattern into a plan.
What it does not do
- It cannot guarantee placement. Assistants are non-deterministic and no one controls their output.
- It is a snapshot, not a trend. A single run cannot separate a real change from ordinary variance.
- It does not game anything. Every recommendation is about publishing better source material.
What happens to your data
Nothing is written to a database, because there is no database. The report lives in the server's memory for thirty minutes and is then dropped, downloaded or not. Logs record timings and counts only — never your URL, the questions or the answers.