The brief
The blocker worth solving is not a website refresh. It is this: UAE buyers increasingly ask an AI, not Google — and we cannot currently benchmark that traffic or operationalise a score against it. G·AI·G turns "are we visible in AI answers?" into a number we measure weekly, own, and can move.
For Azam — this is a peer review, not a finished product. Three questions, straight:
- Is this best practice for GEO / AI-visibility measurement?
- Are these the right tools — or are we missing one you'd reach for?
- Where are the gaps in the method before we operationalise it for GIG?
What we found
Share of answer
83%
GIG cited in 83% of UAE AI answers (50/60, 17 Jun baseline) — strong, but a large slice rides on AXA legacy equity no off-the-shelf tool attributes.
Captured AI traffic
2,336
AI-referred sessions in 28 days (GA4). But 96% is ChatGPT — Gemini, Perplexity, Copilot barely register, because they answer without a click.
What GA4 captures — AI sessions by engine, 28 days
GIG GA4 · "Generative AI Sessions" · 21 May–17 Jun 2026. The bar says it all: GA4 sees almost only ChatGPT, because ChatGPT links out. Every other engine answers without a click — invisible here, but measured by the engine-probe.
The missing-traffic finding. GA4 sees the click; it cannot see the answer. Counting AI-Overview-mediated organic, AI is already driving the equivalent of ~20–28% of the organic channel — and most of it is invisible to analytics. That gap is the reason a deliberate measurement tool has to exist.
The capability gap
Today we measure 3 engines, single-shot, mention-only. To operationalise a score we trust, we add: Perplexity + Gemini (the surfaces we're blind to), n≥3 sampling (confidence, not noise), cited-URL parsing (which sources win), the voices (the named experts AI quotes), and GIG's crawler logs (whether the bots even fetch our pages). Detail on the Tool page; the live numbers on the Report; the build-vs-buy case on Decision.