SEO, AEO & GEO: Understanding the Future of Search

Search used to end with a list of links. Increasingly it ends with an answer — written by a model, assembled from a handful of sources, and delivered without the reader ever visiting a website. That shift has produced three overlapping acronyms, and a good deal of confusion about whether they describe three disciplines or one.

They describe one discipline with three jobs. What follows is what each term actually means, how AI answers choose the sources they quote, and what to do about it in an order that makes sense.

The three acronyms, and what each one means

The terms are used loosely, often interchangeably, and occasionally as though a new one has replaced the last. None of them has.

SEO — search engine optimisation

The original discipline, and still the foundation: making a site technically sound, useful to the person reading it, and credible enough that other sites reference it. Every practice underneath the newer acronyms assumes this work is already done. A page a crawler cannot reach is not a candidate for anything.

AEO — answer engine optimisation

Optimising to be the source an answer is built from, rather than a link ranked beneath one. The unit of success is the citation, not the position. The practical consequence is that an answer engine has to be able to lift a self-contained, correct passage out of a page without needing the rest of the page to make sense of it.

GEO — generative engine optimisation

The same goal, aimed at generative systems specifically — ChatGPT, Gemini, Perplexity, Copilot and the AI results inside traditional search. GEO and AEO overlap almost entirely in practice, and the distinction matters far less than the work they share.

The useful way to hold all three: SEO gets you into the index, AEO gets you into the answer, GEO keeps you there as the models change.

Why the distinction matters now

For most of search's history, visibility and traffic were the same measurement. A page that ranked was a page that got visited, and one number stood in for both.

An AI answer breaks that link. Your business can be described accurately, recommended by name and quoted at length inside an answer that produces no click at all. The visibility is real and the traffic is not, which means the old measurement now reports a loss where there was a gain.

This has two consequences worth acting on. The first is that being cited has value in itself — a reader who arrives already told what you do, by a system they trust, arrives further along than one who found a blue link. The second is that measurement has to change, because a strategy judged only on sessions will conclude that its best-performing work is failing.

How AI answers choose their sources

The systems differ in their details and change often, but the shape of the problem is stable. An answer engine has to find candidate material, decide which of it can be trusted, and lift a passage that answers the question as asked.

That gives three requirements, and they have to be met in order:

• Retrievability — the content can be found and parsed at all.
• Citability — a passage can be lifted out of it and still be correct.
• Authority — the source is one the system has reason to trust.

Work on the third before the first two is wasted. A well-regarded brand whose answers are buried in a page a crawler renders as an empty shell will not be quoted, because there was nothing to quote.

Layer one: make the content retrievable

This is unglamorous and non-negotiable. Most of it is ordinary technical SEO, which is why the foundation matters more rather than less as the acronyms multiply.

• Serve the substance in the initial HTML rather than assembling it client-side after load.
• Keep the crawl path clean: no orphaned pages, no chains of redirects, no blocked resources the page needs to render.
• Mark up what the page is with structured data — an article, an FAQ, an organisation, a product — so a machine does not have to infer it.
• Give every page one clear subject. A page covering four topics is a weak candidate for all four.

None of this is new advice. What has changed is the cost of ignoring it: a page that ranks tenth still exists on a results page, but a page that cannot be parsed does not exist in an answer.

Layer two: make the content quotable

This is where writing for an answer engine genuinely differs from writing for a ranking, and where most content falls down.

A passage that gets quoted answers the question in its first sentence and remains true when read alone. A paragraph that builds to its point over five sentences is good prose and a poor citation, because the sentence a model lifts is the one nearest the question — not the one the paragraph was building toward.

What that looks like in practice

• Ask the question in the heading, in the words a reader would use, and answer it immediately underneath.
• Put the direct answer first and the reasoning after it, not the other way round.
• Keep each answer self-contained — no “as discussed above”, which is false the moment the passage is lifted.
• Define your terms on the page rather than assuming them.
• State plainly what something is not, when confusion between two things is the actual question.

The discipline is closer to writing a good reference entry than a good essay. It also, usefully, makes the page easier for a human in a hurry.

Layer three: build authority as an entity

Models do not weigh a page in isolation. They weigh what is known about the organisation behind it — and what is known comes from consistency across every place the organisation appears, not from the site alone.

• Describe the business the same way everywhere: the site, the directories, the profiles, the press.
• Publish enough depth on a subject that the connection between the business and the subject is unambiguous.
• Earn references from sources that already carry weight in the field.
• Attribute writing to real, identifiable people with real expertise, and say what that expertise is.

This is the slowest of the three layers and the one that compounds. It is also the hardest to fake, which is precisely why it counts.

What to measure

Sessions alone will misreport this work, so the measurement has to widen. Three things are worth tracking, in roughly this order of usefulness:

1. 01Citations: whether AI answers to the questions that matter to your business mention you, and how they describe you.
2. 02Accuracy: whether that description is correct. A confident, wrong summary of what you do is a problem no ranking report will show you.
3. 03Qualified arrivals: what the people who do arrive go on to do, which matters more than how many of them there were.

Checking a fixed set of questions across the major systems on a regular schedule is unsophisticated and works. The point is to have a baseline, so a change is visible as a change.

Where to start

In order, because the order is the whole point: fix retrievability, because nothing else functions without it. Rewrite your most important pages so their answers survive being lifted out. Then build authority patiently, on the subjects you actually want to be known for.

Everything above rests on the same foundation the SEO service page describes, and the AEO & GEO service page covers how we approach the answer-engine layer specifically.

The acronyms will keep multiplying. The underlying requirement has not moved: be findable, be quotable, and be worth quoting.