The search result used to be a list. Ten blue links, and the work was to be one of them.
Increasingly it is a single answer. A buyer asks Google, ChatGPT, Gemini, Perplexity or Copilot for a recommendation, reads one synthesised paragraph, and acts on it. The links are still there, further down, for the minority who scroll. The decision has usually been made above them. If a model cannot read, trust and cite your business, you are not losing the click. You were never in the running for it.
Key takeaways
- Buyers increasingly act on one AI answer, in Google AI Overviews and in chatbots, often without clicking a link at all.
- AI search visibility is won by being legible to a model: structured data, a coherent entity graph and machine-readable content, not by stuffing in more keywords.
- It splits into Answer Engine Optimisation, winning the direct answer, and Generative Engine Optimisation, the authority and entity signals that make a model name you.
- It is measurable. Citations, AI Overview appearances and model recommendations can be tracked rather than assumed.
The front page is now an answer
For twenty years the goal of search was a ranking. A position in a list, earned and defended. The behaviour underneath it was simple: the searcher scanned, chose, clicked, and judged for themselves.
An answer engine removes the scanning and most of the choosing. It reads the sources, decides which to trust, and writes the verdict. Your prospect meets a conclusion, not a menu. That is a different game with a different prize. The prize is no longer being on the page. It is being in the sentence.
A model cannot recommend a business it cannot read.
Why classic SEO is necessary but not sufficient
The technical foundations still matter, because a model cannot cite a page a crawler cannot reach or parse. Clean crawlability, fast pages and correct indexation remain the price of entry. This is the same discipline as technical SEO, and the two are usually run together for exactly that reason.
What changes is the target. Classic SEO earns a position in a list of links. AI search earns a mention inside the generated answer, frequently before anyone clicks anything. You can rank perfectly in the blue links and still be absent from the paragraph above them, because the model assembled its answer from sources it found easier to read and trust than yours.
What a model needs in order to cite you
Three things, layered.
First, machine readability. Structured data that states plainly what you are, what you sell, where you operate and who vouches for you. A model should not have to infer your business from prose; it should be able to read it as fact.
Second, a coherent entity graph. The model is reasoning about entities, your organisation, your people, your services, and the relationships between them. When those are stated consistently across your site and corroborated by trustworthy sources elsewhere, you become a known entity rather than a guess. When they conflict, the model hedges, and a hedge is not a recommendation.
Third, the discovery and hygiene layer underneath: an llms.txt file, clean content the model can quote without distortion, and the speed and crawl health that let it gather all of this in the first place.
AEO and GEO, plainly
It helps to split the work in two.
Answer Engine Optimisation is about winning the direct answer. When someone asks a specific question your business should own, AEO is the structure that makes your content the cleanest available source for that answer.
Generative Engine Optimisation is broader. It is the entity and authority work that makes a model recommend you by name when the question is open, the work that decides whether you are one of the three businesses named or one of the thousand that are not. The two overlap with the schema and entity work in AI search optimisation, and they reward the same thing: a business that is legible and corroborated rather than loud.
How to tell if it is working
The objection to all of this is usually that it cannot be measured. It can. AI Overview appearances, citations inside chatbot answers and the recommendations a model gives for your category can all be tracked, in the same forensic spirit as everything else. You watch whether you are named, for which questions, and how that changes as the entity work lands.
The practice's own site is built this way, and is already cited in AI search. That is the simplest proof of the method: not a promise that it works, but a worked example. If you want the same applied to your business, it starts the way everything here does, with an honest conversation and a forensic look at where you actually stand.
