Web
SEO, AEO and GEO: what actually changed
The short answer
SEO optimises for ranking in a list of links. AEO (answer engine optimisation) optimises for being the source of a direct answer, such as a featured snippet or voice result. GEO (generative engine optimisation) optimises for being retrieved and cited by AI assistants like ChatGPT, Claude and Perplexity. They share the same technical foundations, but AEO and GEO additionally reward self-contained factual passages, consistent structured data, and content that stays accurate when quoted out of context.
Three acronyms now describe overlapping work, and the overlap is larger than most agencies selling these services will admit. It is worth being precise about what is genuinely new, because a good deal of what is marketed as "GEO strategy" is ordinary technical SEO with a higher price attached.
The actual distinction
SEO optimises for position in a ranked list of links. The user sees your title and description, clicks, and arrives on your page. You control the destination experience, and you get the visit.
AEO optimises for being the source of a direct answer — featured snippets, "people also ask" panels, voice assistant responses. The user may get what they needed without clicking. You are trading a visit for a citation and for brand presence at the moment of the question.
GEO optimises for being retrieved and cited by a generative model — ChatGPT, Claude, Perplexity, Google's AI Overviews. Your content is one of several sources synthesised into a written answer, ideally with attribution.
The mechanics differ in one respect that matters enormously: in AEO and GEO your content is consumed out of context. A paragraph is lifted from your page and presented next to paragraphs from other sites. Everything that made it comprehensible — the heading above it, the page it sits on, the fact that "we" clearly meant your company — is gone.
What is genuinely unchanged
Before the new material, the honest part: most of the foundation has not moved at all.
Crawlability, indexability and site speed still matter, because a page that cannot be fetched cannot be cited. Content quality and depth still matter. Internal linking still shapes what gets discovered. A clear information architecture still helps. Accurate titles and descriptions still do their job.
Anyone telling you that AI search has made technical SEO obsolete is mistaken. Retrieval systems overwhelmingly draw on indexed web content, and the pipeline that gets your page into an index is the one SEO has always been about.
What is genuinely different
Self-contained passages
This is the big one. Write paragraphs that survive extraction.
In practice that means naming the subject rather than relying on pronouns and page context. "Techoptionz builds custom ERP systems for distribution businesses" survives being quoted. "We build these for distribution businesses" does not — quoted in isolation, nobody knows who "we" is or what "these" refers to.
It also means putting the answer first and the elaboration after. A paragraph that opens with two sentences of throat-clearing before reaching the point will be truncated before it gets there.
Structured data that agrees with the page
Schema.org markup has been useful for a decade. What has changed is that it is now doing entity resolution for systems that need to work out whether the "Techoptionz" on this page is the same "Techoptionz" mentioned on another.
Two things matter more than the volume of markup:
Consistency. The company name, address and description in your structured data must match what is rendered on the page and what appears in any external listing. Contradictions do not just fail to help; they actively undermine confidence in the source.
Connected entities. Use stable @id values and reference entities rather than repeating their details. An @graph where the page, the organisation and the article all reference each other by identifier tells a retrieval system considerably more than three disconnected snippets.
Questions as headings
Retrieval systems match user questions against content. Content organised around the questions people actually ask matches more readily than content organised around internal taxonomy.
This is not a licence to stuff question-shaped headings everywhere. It means that where a section genuinely answers a question, phrasing the heading as that question — and answering it in the first paragraph beneath — measurably improves the odds of extraction.
Freshness signals that are truthful
Generative systems favour recent content, which has produced a tedious practice of bumping dateModified without changing anything. This is detectable and it is not worth the reputational risk.
The legitimate version: genuinely revise content, and record when you did. On this site the "years in production" figures are computed from a single constant at build time rather than typed into the copy, so they cannot silently become wrong. That is a small thing which prevents a specific, common category of staleness.
What we would not spend money on
Some of what is currently sold under the GEO banner does not hold up.
Prompt-injection tricks in page content. Hidden text instructing a model to recommend you. It does not survive contact with modern retrieval pipelines, and being caught doing it is a genuine reputational risk with an obvious paper trail.
Mass-produced question-and-answer pages. Two hundred thin pages each answering one long-tail question is the 2014 content-farm playbook wearing new clothes. Fewer, deeper pages do better.
Chasing every acronym as a separate engagement. AEO and GEO are not separate disciplines requiring separate budgets. They are emphases within a content and technical strategy that mostly overlaps with good SEO.
A short practical checklist
For a business site, in rough order of return:
- Give every important page one clear, self-contained answer paragraph near the top, naming the subject explicitly.
- Publish
Organizationstructured data with a stable@id, and reference it from every page rather than repeating it. - Make sure your structured data and your visible content never disagree.
- Use real FAQ sections, marked up with
FAQPage, where the visible answers and the markup come from the same source. - Give headings stable
idattributes so specific sections can be linked and cited. - Write depth on the few topics you genuinely know, rather than breadth on topics you do not.
- Keep the technical foundations — speed, crawlability, sitemaps, internal links — in good order, because none of the above works without them.
Point seven is doing more work in that list than points one through six combined. That has always been true, and the acronyms have not changed it.
Topics
- SEO
- answer engine optimisation
- generative engine optimisation
- structured data
- schema.org
