What is SEO?

The three pillars

• Technical. Whether search engines can crawl, render and index the site at all — the foundation everything else sits on.
• Content. Whether the pages answer what people are actually searching for, in enough depth to be the best available answer.
• Authority. Whether other credible sources reference the site, which remains the strongest ranking signal.

What AI answers changed

Search results increasingly begin with a generated answer rather than a list of links. Where the answer is sufficient, the user does not click anything — the query is resolved without a visit.

That has two consequences. Informational queries that used to deliver traffic now frequently deliver none, and being cited within the generated answer has become more valuable than ranking below it.

The response is not to abandon SEO. Technical health, genuine depth and third-party authority are the inputs to both ranking and citation. What has changed is the payoff distribution: the value has concentrated toward being the source rather than being on the page.

SEO and GEO

GEO / AEO is the discipline of being cited or recommended by AI assistants, and it overlaps substantially with SEO without being the same thing.

What transfers: crawlability, structured content, factual specificity and genuine authority.

What does not: keyword volume plays, thin pages produced at scale, and anything optimising for position in a list that no longer exists.

The practical implication is that the half of SEO practice concerned with quality and authority has become more valuable, and the half concerned with volume has become worthless faster than most content programmes have adjusted.

The failure mode worth naming

Publishing many thin pages to cover many keywords was always a weak strategy and is now actively harmful. Search has become better at recognising it, and generated answers do not surface it at all.

A glossary of two-sentence definitions, a set of near-identical comparison pages, or a content farm of AI-generated articles all fail for the same reason: none of them constitute evidence that anyone should be cited. Depth on fewer pages outperforms breadth across many, and by a widening margin.