The best argument for a GEO + SEO programme isn't a pitch — it's a client's traffic graph. So here's one, with the real numbers behind it. whatthewordsays.org is a South African scripture-commentary publisher and a PeakRank client. A year ago it had a small, loyal audience. Today it pulls a five-figure monthly audience and is still climbing.
The numbers, straight from the server logs
These are unique-visitor counts from the site's own AWStats — not estimates. May 2025: 1,439 unique visitors. May 2026: 12,708. That's about 8.8× year-on-year. And the whole of 2025 brought 19,169 unique visitors; by early June 2026 the site had already drawn 38,334 — double the prior year in under half the time.
The content engine didn't change overnight. What changed is that the same good content finally became findable — and citable — across both Google and the AI answer engines people now ask first.
What we actually did
- Fixed the technical foundation — crawlability, speed, headings and structured data on every article, so engines can extract clear, attributable claims.
- Restructured content to be citable, not just rankable — direct answers up top, question-shaped headings, consistent entities. Claude drafts the structure; a human verifies every claim.
- Monitored who the AI engines recommend — Claude reads the ChatGPT, Perplexity, Gemini and AI-Overview answers and extracts who got cited, turning each gap into a brief.
- Turned the site's own data into a weekly to-do list via the per-domain assistant, so the client worked a short, specific list each week instead of guessing.
The honest version: none of this is one big move. It's a few disciplined levers run consistently, with Claude doing the reading and first drafts at speed and a human owning every decision that reaches the client. I wrote up the customer-facing version of this case study on the PeakRank blog — the link's below.