What this can and cannot tell you
What it measures
- Whether assistants name you when asked a question you should own.
- Which competitors take that space instead.
- Whether the gap differs between Arabic and English.
- The most likely reason for each absence, and what would close it.
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.
What actually happens
-
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.