How AI engines choose who to cite — and how to be chosen

What changed

Buyers increasingly ask an AI assistant before they ask Google — and when they do ask Google, an AI answer often appears above the results. AI engines don't rank ten links; they compose one answer and cite a handful of sources. If your business isn't in the sources, you're not in the conversation.

The volume is still small — AI referrals are roughly 1% of most sites' traffic — but those visitors arrive pre-sold: multiple independent studies find they convert at several times the rate of organic visitors. Low volume, high intent. That's a channel worth being early to.

AHREFS: 0.5% OF TRAFFIC → 12.1% OF SIGNUPS · SEMRUSH: 4.4× CONVERSION · DIRECTION CONSISTENT ACROSS 3+ STUDIES

What actually moves citations

Across the credible studies — Ahrefs, Semrush, Profound, Muck Rack, Seer Interactive, the Princeton GEO research — the same levers keep surfacing:

• Classic rankings still carry the most weight. Page-1 Google positions correlate ~0.65 with AI mentions. Good SEO is the foundation, not the enemy, of AI visibility.
• Answer-first structure. 44% of ChatGPT citations come from the first 30% of a page. Self-contained sections that answer one question each get quoted; essays don't.
• Evidence in the copy. Statistics, quotes and named sources lifted citation rates 22–40% in controlled experiments.
• Being talked about. 84% of AI citations are earned media — third-party mentions beat your own website, and mentions correlate with AI visibility far more than backlinks do.
• Fetchability. Pages that fail to load for AI crawlers get ~18× fewer citations. Fast, static, unblocked pages win — most sites block AI crawlers by accident via their firewall.

SOURCES: MUCK RACK 25M LINKS · PROFOUND 1B CITATIONS · SEER · PRINCETON GEO (KDD 2024) · AHREFS 75K BRANDS

What we won't sell you

The AI-SEO market is full of vapor. Things we've tested, read the evidence on, and refuse to sell:

• "AI rankings" guarantees. AI answers vary run to run — over 99% inconsistency in repeat testing, and citation lists churn 40–60% monthly. Anyone guaranteeing a position is guessing.
• llms.txt as a secret weapon. No major engine confirms reading it; a 137,000-site study found 97% of these files were never fetched by an AI bot. We publish one — as a convention, and we say so on it.
• Schema as an AI hack. Two independent studies found no citation lift from structured data. We ship clean schema anyway (it's good hygiene and Bing uses it) — but it's an hour of work, not a monthly line item.
• Blended "AI visibility scores". One number averaging incomparable engines tells you nothing actionable. We track mentions and citations per engine, separately.
• Scaled AI content. Mass-generated pages and fake "best of" listicle farms collapsed in documented waves through 2026. Volume without authority gets filtered.

So what is IEO?

IEO — Intent & Engine Optimization — is our name for the practice that survives the churn: optimising for what people actually mean, and for the engines that now answer them. In practice: rank for the questions (not just keywords), structure pages so machines can quote them, build evidence into every claim, earn mentions on surfaces engines read, and measure per-engine — honestly.

It's not a replacement for SEO. It's SEO, evolved — which, given we've been doing this since 2008, is exactly the kind of evolution we've navigated before.

The Melbourne AI-citation log

From launch, we're tracking who the AI engines actually recommend for Melbourne service queries — monthly, published here. Nobody else measures this for our market. First edition coming soon.

ORIGINAL RESEARCH · MONTHLY FROM AUG 2026