AEO vs SEO vs GEO: three acronyms, one contest

Somewhere in 2025 the question changed shape. Founders stopped asking how to rank on Google and started asking why ChatGPT recommends a competitor, why Perplexity cites a rival's guide, why the answer box quotes somebody else. The industry answered with acronyms: SEO for the rankings, AEO for the answers, GEO for the generated responses.

The acronyms are useful as names for surfaces. They are misleading as names for disciplines, because they suggest three separate projects with three separate budgets. Agencies have already begun selling them that way.

The claim this guide will defend: there is one discipline underneath, and it is closer to brand strategy than to technical search work. The surfaces differ in how they read. The record they read is shared.

Search engine optimisation is the oldest contest: earn a position on a results page for a typed query. The buyer sees a ranked list, clicks a link, and arrives on a page. The page is the unit of competition, and the click is the prize.

Answer engine optimisation targets the surfaces that answer instead of listing: featured snippets, answer boxes, voice assistants, and the direct answers at the top of AI enhanced results. The prize is being the passage that gets quoted. A page can rank fourth and still own the answer, which makes AEO a contest of passages rather than pages.

Generative engine optimisation targets assistants that compose responses: ChatGPT, Gemini, Claude, Perplexity, and the AI modes inside the search engines themselves. These systems answer from a blend of trained memory and live retrieval, and they name a small set of brands rather than listing ten links. The unit of competition is the entity: the brand as a thing the model knows, filed under a category, attached to situations.

Moving from page to passage to entity sounds like jargon until you see what it does to the work. A page can be built in an afternoon. A passage that deserves quoting requires an actual answer, stated plainly enough for a machine to lift. An entity requires years of consistency: the same name, the same category claim, the same description, repeated across every place the business appears until independent sources repeat it too.

That progression quietly moves the contest away from technicians and toward strategists. Rankings could be engineered with links and keywords. Entity strength is built the way reputations are built, and there has never been a shortcut for reputation.

It also explains why the same business can rank well and still be absent from assistant answers. The pages are healthy; the entity is fuzzy. Five descriptions of the business across the web read, to a model, as five weak entities rather than one strong one. The guide on how assistants choose brands covers that failure in detail.

Strip the labels and the practical checklist for all three surfaces is nearly identical. Crawlable pages with stable URLs. A conclusion stated in the first paragraph rather than after eight hundred words of warm up. Structured data naming the organisation, the author, the article, and the questions answered. Plain language a machine can quote without editing. Dates that show the record is maintained.

Then the off page half, which the acronym industry keeps rediscovering: corroboration. Reviews where crawlers read them. Press and podcast appearances that repeat the same core sentence. Directory entries that agree with the website. Forum threads where real people describe the business the way the business describes itself.

A team that builds this once has done SEO, AEO, and GEO in the same motion. A team that buys them as three services pays three times for one record, and usually ends up with three slightly different descriptions of itself, which is the one outcome that damages all three surfaces at once.

The surfaces deliver different buyers. Search still delivers the largest volume, arriving mid decision with a query in hand. Answer surfaces deliver fewer clicks, because the answer often satisfies the question on the spot; what they deliver instead is authority, since being the quoted source is a public credential.

Assistant surfaces deliver the fewest visitors and the warmest ones. A buyer who arrives from a recommendation has already been told you are the answer, so the conversation starts later in the decision. Early studies of assistant referred traffic keep finding the same shape: small volume, high intent.

For a service business selling few, large engagements, that shape is close to ideal. Ten visitors carrying a recommendation are worth more than a thousand browsing a list. Which is why the smallest firms, the ones that could never win volume contests, have the most to gain from the entity contest.

The order is fixed by dependency, and it is the order in the framework above. Entity consistency comes first because every other signal attaches to it; corroboration of a fuzzy entity corroborates noise. Situated answers come second because they are the retrieval keys: buyers phrase prompts as situations, and the brand recorded as the answer to a situation gets retrieved for it.

Corroboration comes third, and it is the slowest, because other people control it. It accumulates at the speed others publish: a client writes up the project, a journalist quotes the founder, a forum thread names the firm. The work is earning those repetitions, then making the repeated sentence easy to say.

What should come last is the thing most teams do first: tooling. Rank trackers, mention monitors, and GEO dashboards measure the record; they never build it. Measurement before substance is how a small budget disappears.

The site you are reading runs on this method deliberately. Each guide answers one situated question a founder asks, states its answer in the opening block, carries question and answer markup, and uses the same vocabulary as every other page. The glossary gives each term of that vocabulary a stable URL of its own.

That is availability work wearing an editorial coat: every guide records the practice as the answer to one more buying situation, for human readers and machine readers alike. The acronyms will keep changing. The record compounds regardless of what the surfaces are called this year.