Generative Engine Optimisation (GEO): Getting Cited by AI

Ask ChatGPT to recommend a partner in your field. Ask Perplexity what your company does. Ask Google’s AI Overview who the experts are. Whatever those answers say, that’s the new shop window, and most businesses have no idea they’re invisible in it.

Search is splitting in two. There’s still the ranked list of blue links. But increasingly there’s an answer, written by an AI that read the web and synthesised a response, often with a handful of cited sources and no need to click anything. Getting into that answer is a different game from ranking a page. That game is Generative Engine Optimisation.

This guide is the playbook, and a live case study: this very site is built to be cited by AI, and we’ll show you exactly how.

What GEO actually is

Generative Engine Optimisation is making your content easy for AI answer engines to read, trust, and cite. The term comes from a 2024 Princeton and Georgia Tech paper that tested it systematically across 10,000 queries and found that targeted moves, citing sources, adding statistics, including direct quotations, boosted a source’s visibility in AI answers by up to 40%.

You’ll also see it called AEO (Answer Engine Optimisation). Same idea, two names. The distinction that matters is from SEO:

	SEO	GEO / AEO
Competes for	A ranked link	A citation inside the answer
The user gets	A list to click	A written answer, maybe a link
Optimises for	Crawlers + ranking signals	Readability, trust, extractability
The win	You’re on page one	You’re the source the AI quotes

These aren’t rivals. Most GEO work, clean content, real expertise, structured data, also helps you rank. GEO is a layer on top of good SEO, for a surface that didn’t exist three years ago.

Why it matters now

People are asking AI assistants the questions they used to type into Google, and acting on answers that name a few sources and skip the rest. If your business isn’t in that synthesis, you’re not on page two; you’re not in the conversation. And unlike the ranked list, there’s often no second result to scroll to, the answer names who it names.

For a services business, the stakes are concrete: when a prospect asks an AI “who should I talk to about getting AI into production,” you want to be a name it can confidently give, with a reason attached.

The GEO playbook (and how this site does it)

Here’s the practical work. We’ll point at where this site does each one, so you can verify it, not just take our word for it.

1. Let the AI crawlers in. Answer engines use named crawlers, GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended. If your robots.txt blocks them (many sites do by default, or by accident), you’ve opted out of being cited. Ours explicitly welcomes them, see for yourself.

2. Publish an llms.txt. A model’s context window is too small to crawl your whole HTML site, so you hand it a clean map instead. llms.txt is an emerging standard, a markdown file at your root that indexes your content with links to plain-text versions. Here’s ours.

3. Give them machine-readable pages. HTML is full of navigation, scripts, and styling an LLM has to wade through. We publish a plain-markdown twin of every page, add .md to any URL on this site and you’ll get the clean version a model would rather read. It’s the difference between handing someone a printed page and a tangle of cables.

4. Mark up your entities and answers with schema. Structured data (JSON-LD) tells an AI, unambiguously, who you are (Organization), who wrote this (Person, with real credentials), and what questions this page answers (FAQPage). It removes the guesswork from “is this a credible source on this topic.” Every page here carries that entity graph, the production-ready AI guide and the others all ship FAQ + author + organisation schema.

5. Write answers an AI can lift. This is the content half, and it’s where the Princeton findings bite. Lead with a direct, quotable definition. Use statistics with sources. Add an FAQ. Make claims an engine can extract and attribute without ambiguity. Vague, throat-clearing prose ranks badly and cites worse.

6. Be a real entity with real expertise. Answer engines weigh authority. A named author with a track record, consistent mentions across the web, and citable, first-hand experience all make you a safer source to quote. There’s no shortcut here, and that’s the point, it favours people who actually know the work.

How to measure GEO

GEO is harder to measure than SEO, because the “impression” happens inside someone else’s model. A few practical signals:

• Ask the engines directly. Query ChatGPT, Perplexity, Claude, and Google’s AI mode with the questions your buyers ask. Are you mentioned? Cited? Described accurately? Re-check monthly.
• Watch AI referral traffic. Visits from chatgpt.com, perplexity.ai, and gemini.google.com are a growing line in analytics, segment them.
• Track AI-Overview impressions in Search Console as Google expands its coverage.
• Monitor brand mentions, not just links. In an answer-engine world, an uncited mention still shapes a buying decision.

Who does this work

GEO isn’t a plugin you install. It’s the same discipline as the rest of good engineering: clean structure, honest content, real expertise, made legible to a new kind of reader. We build it into everything we ship, because we had to do it for ourselves first, this site is the proof.

If your business is invisible in the answers your customers are already reading, that’s worth fixing early, while the surface is still new and uncrowded. Tell us what you’d want an AI to say about you, and we’ll work back from there.