# GIG Content Audit — how it was done, so you can re-run it

**Prepared by Articulate AI · 2026-06-17.** The opposite of a black box: here is every data source, every script, and every step. Nothing here is proprietary to us — you own it and can re-run it.

**The data lives in Airtable:** `GIG Content Audit` base — https://airtable.com/app8ohZdWxcXwy2yg
Four tables: **Audit Register** (33 audits: status · finding · how-to-re-run · data link), **Blog Pages** (147, GSC 16mo), **Product & Funnel Pages** (43, GSC 16mo), **AI Visibility** (60, this week's run).

---

## 1. Search Console data (the performance spine)

1. Search Console → giggulf.ae → Performance → Search results → date range **16 months**.
2. **Export → Google Sheets.** You get tabs: Queries, Pages, Countries, Dates.
3. To slice without the 1,000-row UI cap, query the sheet directly. Example — all blog pages by clicks:
   `https://docs.google.com/spreadsheets/d/<SHEET_ID>/gviz/tq?tqx=out:csv&sheet=Pages&tq=select A,B,C,E where A contains '/blog/' order by B desc`
   Swap the filter for `/personal/products/` (product pages) or `not A contains '/blog/'` (everything else).

**What it told us:** 306K clicks / 16mo · 80% brand demand · Motor/Utility 897 clicks/post vs Travel 86 · top-5 posts = 68% of blog clicks · traffic-fines page 1.07M impressions at 0.64% CTR · 37% of posts under 10 clicks.

## 2. GA4 (the conversion layer)

GA4 → giggulf.ae property (it's on the gig-gulf Google account, `authuser=1`) → Reports → Engagement → Pages and screens → set 12 months → export. Multi-market: UAE, Bahrain, Qatar, Oman all report. Conversion anchor: car-insurance click-to-quote ≈ 4.24%.

## 3. Full estate map (the taxonomy)

Firecrawl sitemap of giggulf.ae → 1,765 URLs. Group by path: document library 527 · blog 135 · Arabic 81+ · product/funnel 25 · business/claims/tools 37 · ~940 listing/utility. Same tool maps competitors (Sukoon, Salama, Orient, Lemonade) for the content-volume comparison.

## 4. AI-visibility tool (the GEO measure) — `AIVisibility/`

The in-house replacement for a $3.5k–9.5k/month agency dashboard. Three files:

- **`prompts.json`** — the UAE buyer prompt set, tagged by funnel phase (Category / Attribute / Competitive / Trust / Utility) and LOB. Edit/extend freely.
- **`run.py`** — asks every prompt across ChatGPT, Claude and Google, detects whether GIG (incl. AXA legacy equity) and competitors are cited, saves a dated snapshot to `snapshots/`. Re-run: `python3 run.py`.
- **`report.py`** — turns the latest snapshot into an **action-first** weekly report (`weekly-report.html`): citation share by phase, competitor leaderboard, and a ranked "do these this week" list. Re-run: `python3 report.py`.
- **`load_airtable.py`** — pushes the data + the audit register into the Airtable base.

**Weekly cadence:** `python3 run.py && python3 report.py`. Next week shows movement — what we won back, where competitors moved. Add a Perplexity or Gemini key to widen the engine set.

**Keys** are read from the Articulate credentials file; to run on GIG infrastructure, drop your own OpenAI + Anthropic + Serper keys into the three `key(...)` lookups at the top of `run.py`.

## 5. What's done vs queued

See the **Audit Register** in Airtable for live status. Done: estate map, GSC pull + all the performance analysis, GA4 access, AI-visibility baseline, product-page + out-of-/blog/ analysis, competitor crawl, the database. Queued (need more time or access): full 131-post text scrape + richness scoring, the 527-doc library, schema pass, multi-market GSC, backlinks (Ahrefs tier). Each row says exactly how to finish it.