Every few years something arrives that is supposed to kill SEO. This time the claim has more substance than usual, not because search is dying, but because the unit of visibility is changing. For twenty years the goal was a position in a list. Increasingly, the goal is to be the source an answer is built from.
That shift has a name: Generative Engine Optimization (GEO). It is worth being precise about what actually changes, because most of the advice being written about it is either recycled SEO or wishful thinking.
What stays exactly the same
Start here, because it is the part people get wrong in their enthusiasm. Generative engines are not conjuring answers from nothing. They retrieve, and then they summarise. That means the retrieval layer still matters enormously:
- Crawlability. If a crawler cannot fetch and parse your page, you cannot be cited. This is not negotiable.
- Being indexed somewhere that matters. Perplexity and ChatGPT's browsing lean heavily on conventional search indexes and a handful of high-authority sources.
- Topical authority. Models and retrievers alike favour sources that cover a subject in depth rather than once, shallowly.
If your technical SEO is broken, GEO is not a strategy. It is a distraction. Fix the foundation first.
What genuinely changes
1. There is no position ten
A blue-link SERP has ten slots and a long tail of partial credit. An AI answer typically cites between one and five sources. The distribution is brutally top-heavy: you are cited or you are invisible. There is no "we rank seventh and still get some traffic" consolation prize.
2. Passages get cited, not pages
Classic SEO optimises a page against a keyword. Generative engines pull passages. A 3,000-word guide that buries the answer under an anecdote will lose to a page that answers the question in a clean, self-contained paragraph, even if the long guide is objectively better researched.
Write so that any single section can be lifted out and still make sense on its own. That is the single highest-leverage change most teams can make.
3. Being quoted is not the same as being recommended
This is the trap that catches most brands measuring their own AI visibility. A model mentioning your name is not a win if the sentence is "unlike X, which lacks…". Naive brand-mention counting will happily score that as a citation. Any measurement worth trusting has to distinguish a recommendation from a passing reference, or worse, a mention inside a denial.
4. The query is a conversation
Nobody types "best crm software" into an assistant and stops. They ask a question, then narrow it: budget, team size, integrations. Your content has to survive the follow-ups, which means covering the qualifying details (pricing model, limits, who it is not for) that classic SEO copy tends to hide.
The practical shortlist
- Answer in the first paragraph. Then elaborate. Inverted pyramid, like a newsroom.
- Use real headings as questions. They map directly onto how people prompt.
- Be concrete and falsifiable. Numbers, dates, named limitations. Vague marketing prose is unquotable.
- Say who it is not for. Models reward sources that read as balanced, and it earns you the comparison queries.
- Keep facts current. A confidently wrong date is how you get dropped from an answer set.
So is SEO dead?
No, but "rank for the keyword" is no longer the whole job. The honest framing is that GEO is a second surface layered on the first: the same crawlable, authoritative content, restructured so a machine can lift a clean answer out of it and attribute that answer to you.
The teams doing well here are not doing anything mystical. They write clearly, structure deliberately, and, critically, they measure whether the engines actually cite them, rather than assuming.
