# GIG Gulf — AI-Visibility: position, provider landscape, and the data we don't have
_Brief prepared 2026-06-18. Baseline data: AIVisibility/snapshots/2026-06-17.json (20 prompts × 3 engines, UAE)._

## 1. GIG's current AI-visibility (the "audit")

Counting AXA legacy as GIG-owned (correct — AXA Gulf is the pre-rebrand entity, ~25% of GWP rides on AXA intent):

- **Overall citation share: 83% (50/60 answers).**
- By engine: Claude 100% (20/20) · Google 80% (16/20) · ChatGPT 70% (14/20).
- By phase: Competitive 100% · Trust 100% · Category 92% · Attribute 83% · **Utility 33% (weak).**
- By LOB: Travel / Home / SME 100% · Brand 89% · Health 83% · **Motor 73% (weakest, biggest book).**
- Competitor leaderboard (appearances of 60 answers): Sukoon 23 · Orient 14 · Daman 13 · Yallacompare 12 · ADNIC 11 · RSA/Liva 8 · Dubai Insurance 7.

**Load-bearing caveat:** any external tool (GeoEasy, Zainlee, Profound) matching only the literal string "GIG Gulf" will understate visibility and report a false category gap that AXA equity already fills. Our tool disambiguates AXA→GIG; no off-the-shelf tool does.

## 2. Provider landscape

- **UAE-visible self-serve:** Zainlee (zainlee.com) + analyser at aigeo.ae. Rest of UAE layer is agencies (Zenerom, PUSH, Tenet, Crescendo).
- **Global tools that matter:** Profound (leader; front-end capture 10+ engines + consumer panel + CDN-log crawler analytics; $99–499+/mo), Semrush AIO ($120+), BrightEdge (entity/KG, custom), AthenaHQ (schema/entity automation, $49–295), Otterly ($39), Peec AI (~$90), Scrunch (enterprise), KAI Footprint (free/$500+), Gumshoe (persona modelling, beta), Writesonic/Addlly (content gen).
- For GIG the only two worth a look: **Profound** (data moat, regulated-industry posture) and **Zainlee** (local).

## 3. Data GIG does NOT have — even with our tool running

Closeable on our own stack:
1. **Engine coverage + real surfaces.** Today: gpt-4o-mini API, Claude Haiku API, Google SERP. GIG is unmeasured on Perplexity, Gemini AI Overviews, Copilot, DeepSeek; ChatGPT/Claude figures are model-API, not live-browse/SearchGPT answers.
2. **Variance + cited-source URLs.** Single run = noisy (no repeated sampling, no confidence band). Tool logs a mention, not which URL the engine cited (giggulf.com vs aggregator vs competitor page).

Moats GIG can only license, not build:
3. **Real prompt-volume / demand data.** Our 20 prompts are a guess. No data on actual query volume by topic. Profound's opted-in panel (400M+ conversations, region/demographic) is unreplicable.
4. **AI-crawler / server-log data.** Whether GPTBot/ClaudeBot/PerplexityBot/OAI-SearchBot fetch giggulf.com pages, and which. Needs GIG CDN logs. Tells you *why* you're cited or not.
5. **Sentiment + position-in-answer.** We record yes/no only. No rank-in-answer, no favourable/unfavourable read — material for an insurer (a "claims complaints" citation is a visible liability, not a win).

## 4. The line for the GIG conversation

GIG's AI visibility is already strong (83%); weak seams are precise and fixable (Motor, Utility how-tos, ChatGPT). The two signals that would actually move strategy — what buyers really ask, and whether bots crawl the site — are the data we don't have and can't fake. Position: "we measure your share-of-answer better than any tool because we handle AXA equity; a Profound-class license would add demand data and crawler logs on top."

## Next moves (pick)
- [ ] Extend `aiviz-run.py`: add Perplexity + Gemini, repeated sampling (n≥3), parse cited URLs.
- [ ] Request GIG CDN/server logs to measure AI-bot crawl of giggulf.com.
- [ ] Add sentiment + position scoring to the report.
- [ ] Re-run weekly; turn the 17 Jun baseline into a trend.