AI engines don't browse — they read.

The mechanism, plainly stated

Google ranks pages. AI engines read them — and then decide whose words are worth quoting.

Three things decide who gets cited:

Structure the machine can parse. GPTBot, ClaudeBot, and PerplexityBot do not execute JavaScript. If your critical content — your headings, your service descriptions, your proof — exists only inside a client-rendered JavaScript shell, these crawlers see an empty page. Your content doesn’t exist to them. This is documented behaviour, confirmed in Vercel and MERJ’s December 2024 crawl study of 10,000 production URLs.¹

You can check this on your own site in about ten seconds, free and without signing up: the AI crawler check fetches your homepage as each assistant’s crawler and reports which ones can actually read you. Your robots.txt alone won’t tell you — a server or CDN rule can block a bot that the file appears to allow.

Statistics. The Princeton GEO study (KDD 2024) measured what moves a page’s share of an AI answer. Adding statistics lifted visibility by 31%, on pages the engine had already chosen as sources.²

Attributed quotes and outbound citations. In the same test, adding quotations lifted visibility by 41%. Citing sources lifted the fifth-ranked page by 115%, while the top-ranked page lost 30%.²

It is a lab study, run mostly on a simulated engine, not a promise about your rankings. We build to it because a page that carries checkable facts and names its sources is worth quoting either way.

The context behind the shift

Around 48% of Google queries now trigger an AI Overview.³ On those queries organic click-through runs at 2.36%, against 3.82% on queries without one, and a page cited inside the Overview gets 120% more organic clicks per impression than one that isn’t.⁴ The traffic behaviour of the web is changing — the question is whether your content structure is ahead of it or behind it.

We hold ourselves to the same read. Our own engine audit (AAA, 2026-09-27, audit engine v3) returned GEO 70/100 and SEO 95/100. We publish our own read, weak rows included, from the same engine that will read yours.

How the work is delivered

There is no fixed-scope SEO/GEO product. The work is priced by what the audit reveals, not by a price list.

Read → fix path. The free scan runs nine domains including GEO and content structure, and The Full Read goes further, into your claims, your terms and who four AI models name when your buyers ask. Remediation tiers (R3,000–R65,000) scope and execute the fix work. The read prices the work; you don’t commit to a fix before you know what’s broken.

Retainer cadence. For ongoing GEO authority — two to three pages per month built to the citation floor (≥3 statistics, ≥1 attributed quote, ≥3 outbound citations per ~1 000 words) — this is available inside the Monthly Advisory Retainer.

No structures are built to pass an audit metric. The discipline is what gets you cited; the audit measures whether you have it.

What we don’t do

No off-page link-building. No PBNs. No bulk guest posts. No paid backlinks. The four whitelisted outbound tactics — broken-link swap, named-publication guest posts, HARO, and paid premium editorial placements — are the ceiling.

No FAQ schema. Google deprecated FAQ rich results on 7 May 2026. Emitting FAQPage JSON-LD is now a finding in the audit, not a signal. Visible Q&A content is fine; the schema is not.

The read comes first

The free scan produces a scored read across nine domains. It tells you where the machine sees gaps before any work is scoped. When the question is bigger than the website, The Full Read reads the whole business.

The methodology behind GEO and the dual-reader thesis is on the AI Visibility page. This page carries the floor and links there — it does not attempt to be the evidence page.