LLM SEO and AI Search: What Actually Gets You Cited

LLM SEO, generative engine optimisation, answer engine optimisation. Three names, one problem: when someone asks ChatGPT, Claude or Gemini who to hire in your category, do you get mentioned, and is what it says about you correct?

That is worth solving. It is not worth buying three times, which is what the market is currently selling. We set out the commercial side on our AI SEO page. This post is how the thing actually works.

Ranking and being recommended are different jobs

Google ranks pages. It decides which ten links to show and in what order, and you either appear or you do not.

A language model does something else. It reads, and then it writes an answer in its own words, naming a few sources if you are lucky. You are not competing for a position. You are competing to be the thing worth quoting.

Those reward different work. You can hold page one and be absent from every AI answer in your category, which is the situation most businesses are in without knowing it.

What actually makes a model quote you

Answers that come first. Models lift passages that answer a question directly. A page that spends four paragraphs warming up before saying what something costs loses to one that says it in the opening sentence. This is the single cheapest change most sites can make, and almost nobody makes it.

Structure a machine can read. Structured data states plainly what your business is, where it operates and what it sells, instead of leaving a crawler to infer it from marketing copy. Missing or contradictory markup is the most common reason a real business reads as thin.

Corroboration you did not write. This is the one that separates AI search from ordinary SEO. If everything findable about your business was published by your business, you look unverified. Independent mentions, a complete Google Business Profile, reviews, and local citations turn a claim into something a model will repeat.

Details that agree with each other. One business name, one address, one phone number, everywhere. A stale directory listing with an old number can quietly become the source a model trusts over your own site.

People ask questions now, not keywords

The shift that has been building for years and is now finished: nobody types "SEO Halifax" into an AI tool. They ask "who is the best SEO agency in Halifax?" or "should I hire an agency or do this myself?"

Those are whole questions with context in them, and they get whole answers back. It is why an FAQ section written properly outperforms a page of keyword-optimised prose for this kind of visibility. You are not matching a phrase. You are supplying a usable answer to a question someone actually asked.

The practical version: write down the questions buyers ask you on calls, and answer them on the page, in plain language, the question as the heading and the answer directly underneath. That is most of the work.

Answers are personal, so there is no single ranking

Two people asking the same AI tool the same question can get different answers, shaped by their location, their phrasing and the conversation they were already having.

That matters for how you judge progress. There is no position three to celebrate. The honest measure is whether you appear at all, across several phrasings, on more than one model. Check it yourself: open ChatGPT, Claude and Gemini, ask who the best option is for what you do in your city, and write down what comes back. That is your baseline, and it costs ten minutes.

What is genuinely new, and what is not

Not new: technical health, clear writing, a site that loads, knowing what your customers actually want. That work was always the job.

New: corroboration matters more than it did, because a model weighs what other sources say about you more heavily than a search engine does. And plain answers beat clever copy, because the model is looking for something to quote rather than something to rank.

What is not true, despite being widely sold: that AI tools automatically fix your technical SEO, predict what customers will search before they do, or write content that ranks because it was machine-made. We have seen all three claimed. None of them survive contact with a real site.

Where to start

Ask the three models what they say about your business. If the answer is wrong, thin, or describes a competitor, you have found your problem and it is solvable.

Then look at your own pages and ask whether a stranger could get a straight answer from the first two sentences of each one. Most of the time they cannot, and fixing that helps ordinary search at the same time.

If you would rather have it looked at properly, the AI Visibility Check runs the question across all three models and documents the answers. If you want the ongoing search work with this handled inside it, that is what we do.

Frequently asked questions

What is LLM SEO?

LLM SEO is the work that decides whether large language models like ChatGPT, Claude and Gemini mention your business when someone asks for a recommendation. It overlaps heavily with SEO. The genuine difference is that models weigh independent corroboration, what other sources say about you, more heavily than search engines do.

Is LLM SEO the same as GEO and AEO?

Yes, in practice. Generative engine optimisation, answer engine optimisation and LLM SEO are three labels for one problem: whether an AI tool recommends you correctly. They appeared within about eighteen months of each other because every agency needed a name for it. Treat anyone selling them as three products with caution.

How is AI search different from Google ranking?

Google ranks pages and shows you links. A language model reads sources and writes an answer in its own words. You are not competing for a position, you are competing to be worth quoting. That is why you can hold page one on Google and appear in no AI answers at all.

How do I find out if AI mentions my business?

Ask it. Open ChatGPT, Claude and Gemini and ask who the best option is for what you do, in your city. Try two or three phrasings, because answers are personalised and there is no single ranking to check. Write down what comes back so you have a baseline to compare against later.

Does AI write content that ranks?

Not on its own. Machine-written content ranks the same way any other content does, on whether it answers the question better than the alternatives. What AI is genuinely useful for is speed of drafting and finding gaps. What it will not do is supply the judgement, the specifics or the corroboration that make a page worth citing.

What should I fix first for AI search?

Put a direct answer in the first two sentences of every page, and make sure your business details are identical everywhere they appear. Those two cost nothing and fix the most common reasons a model skips or misdescribes a business. Structured data and corroboration come next.