GEO measurement — the case, where we stand, and the options

Why we need this

UAE insurance buyers increasingly ask ChatGPT, Perplexity and Gemini — not Google — "what's the best car insurance in Dubai?" or "how do I check my black points?". If GIG isn't named in that answer, it's invisible at the moment of intent, and a competitor takes the lead.

As Performance Director, GEO recommendations have to stand on data we own and can re-run — not a vendor's opinion. That means measuring, every week, whether GIG wins the AI answer, where it loses, and to whom.

What we can do today

We already run aiviz — our own tool — across ChatGPT, Claude and Google, AXA-aware, weekly. Current baseline (17 Jun):

• GIG cited in 83% of UAE AI answers (50/60).
• Strong on Category, Competitive, Trust; weakest on Motor utility (government how-tos), where Daman & aggregators win.
• Capability today: a working measure we own — roughly 5/10.

Where it falls short

Good enough to spot a gap; not yet good enough to defend a board-level recommendation:

• One run (n=1) — can't tell a real gap from a bad day.
• 3 engines — Perplexity, Gemini, Copilot not measured.
• Mention only — not which URL is cited, not sentiment.
• Prompts are guessed — no real demand data behind them.
• No crawler truth — blind to whether AI bots even fetch GIG's pages.

The options