AI search optimisation · GEO · AIO · LLMO
When your customers ask ChatGPT, be the answer it gives.
AI Overviews, ChatGPT and Perplexity answer the question and credit a few sources. This page explains how they choose, and what I do to make your pages one of them.
How the engines pick their sources
Each one works a little differently and the differences matter. Google's AI Overviews and AI Mode split your question into several smaller searches, pull passages from the pages that rank for each, and write an answer that cites the passages it used. Ranking still counts, but it's the passage that gets picked, not the page. ChatGPT runs a live search for anything recent, leaning on Bing's index and its own crawler, and favours pages that answer directly in the first few hundred words. Perplexity keeps its own index, weights a list of domains it already trusts, re-ranks for quality and watches how a new page performs in its first days. I pulled the Perplexity mechanics apart for AI Marketing School.
What a citable page looks like
A heading that matches the question as people ask it. A direct answer in the first paragraph, in a sentence a model can lift whole. Figures with a named source and a date. A definition that stands on its own. Short sections, one idea each, with FAQ and Article schema so the structure is explicit. And the crawlers let in. GPTBot, OAI-SearchBot, PerplexityBot and Google-Extended are blocked by a surprising number of robots.txt files, usually by accident.
Why you can rank and still not be cited
Because the engines quote a passage, not a position. A page that ranks third but buries the answer in paragraph nine loses to one that ranks eighth and answers in paragraph one. Vague copy loses to specific copy every time. And a brand nobody else mentions loses to one that turns up on Reddit, in the trade press and on the comparison sites the models read. Citations follow mentions.
Getting cited by AI is the new link building, but AI systems need extractable, attributable information they can quote and reference.
AI Marketing School, "The great unclick", June 2025
Plan explosive early distribution within hours of publishing.
AI Marketing School, "Gaming AI search like it's 1999", August 2025
Understanding this basic principle might help you get great results with LLMO.
AI Marketing School, August 2025
The offer
The work
No plugin and no dashboard. Editorial work, from the first audit to the sentence a model eventually quotes.
AI visibility audit
I run the questions your buyers ask through ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews, log every source cited, and show you who is winning and with which passages. A short report, not a deck.
Citable content
New pages and rewrites built around the sentence a model will lift, with the answer up front and the sources named.
Structure and markup
Headings that match real questions, FAQ and Article schema, internal links that make the page's subject obvious, and a robots.txt that lets the right crawlers in.
Authority and distribution
A launch plan for each piece, so it reaches real readers in its first hours, and placements on the sites the models already draw on.
Brand in the corpus
Your name, your definitions and your point of view on the sites that feed the models, so you show up even when your own page doesn't.
Monthly tracking
The same questions re-run each month, so you can watch citations appear and see which pages earned them.
Straight talk
Ethical growth hacking
Every ranking system gets gamed and this one is no exception. Hidden instructions, prompts tucked into alt text, that sort of thing. It works for a fortnight and then the domain disappears. I don't do it.
What I do use are the levers that last. Early distribution in the first hours, when the engines are deciding what a new page is. Exact-match titles on questions nobody has answered yet. Your name and your definitions on the sites the models already trust. Quiet questions take weeks to move and busy ones take months, and I'll tell you which you're facing before we start. None of it replaces good writing. Models quote the sentence that was worth quoting.
Find out who AI cites instead of you
Send me your site and the five questions your customers ask most. I'll run them through the main AI tools and send you an honest view of where you stand.
Request an audit →Questions people ask
Questions about AI search
What is AI search optimisation?
Also called GEO, AIO or LLMO, it's making your content the source that ChatGPT, Perplexity, Claude, Gemini and Google's AI Overviews cite when they answer a question. It overlaps with SEO, but it rewards specific, sourced statements, early distribution and a presence on trusted domains.
How is GEO different from SEO?
SEO gets you ranked in a list of links. GEO gets you quoted inside the answer. AI engines cite several sources at once, prefer specific claims to vague ones, lean towards trusted domains and notice early engagement. Good SEO content is a strong start, but it usually needs restructuring to be cited.
How long does it take to get cited by AI search?
Weeks for quiet questions on a site with some authority, a few months for busy ones. I audit first and tell you which you're facing.
Can AI search be gamed?
Yes, and I've written up how. The crude tricks, hidden text and injected instructions, work briefly and then cost you the domain. The durable ones are early distribution, exact-match titles on unanswered questions and a presence on the sites the models trust. Those I use without apology.
Do I need an llms.txt file?
It takes ten minutes and does no harm, but none of the major engines has said it uses one. Letting the crawlers in through robots.txt and structuring the page properly matter far more.
What does an AI visibility audit include?
I run the questions your buyers ask through ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews, log every source cited, compare them with your pages and send a short report on what would close the gap.