Every few years the marketing industry declares SEO dead, and every few years SEO fails to die. So when you hear that generative engine optimisation (GEO) is "the new SEO", some scepticism is healthy.
The truthful answer is less dramatic and more useful: AI search changes where the buying decision happens, but the machinery underneath it still runs on the same web that SEO has always optimised.
This article lays out the real differences without the hype: what carries over from SEO, what genuinely changes when a language model writes the answer, and what the research says about which work actually moves the needle.
What stays the same (more than you'd think)
AI engines don't conjure answers from thin air. When ChatGPT or Gemini answers a commercial question, it typically retrieves live web content first, and that retrieval step looks a lot like classic search.
Pages that are crawlable, fast, well-structured and genuinely relevant get retrieved. Pages that are blocked, bloated or thin don't.
If your site has technical SEO problems, AI engines inherit every one of them.
That means the unglamorous fundamentals (clean site architecture, indexable pages, descriptive headings, real content depth, authoritative links) are not the "old way". They are the supply chain for AI answers.
A business that skips them and goes straight to "AI optimisation" is decorating a shop the delivery trucks can't reach.
What genuinely changes
The differences are real, and they cluster around one fact: the output is a written answer, not a ranked list. That reshapes the job in four ways.
- From ranking to being cited. In classic search, position three still gets clicks. In an AI answer, you're either named or you're absent. The unit of success shifts from rankings to citations and mentions.
- From keywords to questions. People type fragments into Google but ask AI full conversational questions: "which mortgage broker in Parramatta is good for first-home buyers?" Content has to answer questions, not just contain keywords.
- From your site to the whole record. Engines weigh what independent sources say about you (reviews, directories, media) more heavily than search ever displayed. Your citability depends on your entire public footprint, not just your domain.
- From clicks to influence. Some AI answers resolve without any click at all. Measurement has to follow the effect (branded searches, direct enquiries, "heard about you from ChatGPT") rather than counting sessions alone.
The uncomfortable evidence about AI-copy makeovers
Here's the caveat an honest agency has to give you. Controlled research on generative engine optimisation keeps finding the same pattern: whether your content gets retrieved at all (a technical SEO and relevance problem) matters more than cosmetic rewrites of the copy.
Sprinkling statistics or authoritative-sounding phrases into pages that engines never retrieve moves nothing. The foundations beat the makeover.
1.82×
referral-traffic lift estimated in a controlled AEO field study
That lift, from a controlled field study of answer-engine optimisation, is what disciplined work can produce. Note what it measures: referral traffic, not revenue.
The gains show up when structural work, retrieval and third-party authority move together. They should always be verified against controls in your own market rather than taken on faith.
So is it GEO or SEO? It's an order of operations
The framing we use with clients: GEO extends SEO, it doesn't replace it. Lead with GEO because AI answers are where shortlists are now written, but keep SEO foundational, because retrieval decides what the AI even sees.
In practice a single programme covers both: technical foundations and content depth serving Google rankings and AI citations at once, plus the GEO-specific layers of entity consistency, third-party authority and answer-ready structure.
A concrete example makes the sequence tangible. Take a suburban accounting firm whose site loads slowly, hides its pricing and has three name variants scattered across old directory listings.
Rewriting its homepage "for AI" achieves nothing, because engines struggle to retrieve the site and can't tell which listing is really the firm. The right order is: fix the technical issues so pages get retrieved, publish a straight-talking fees page and a proper services breakdown, reconcile every listing to one identity, then build reviews and local mentions.
Each step is classic SEO or classic reputation work. Together they're what gets the firm named when someone asks an AI for an accountant nearby.
Where to put the next dollar
If your technical SEO is broken, fix that first: it's the highest-return GEO work there is. If your foundations are sound, the next dollar goes to decision-useful content and third-party proof, because that's what converts retrieval into recommendation.
Our services are built around exactly that sequence, and our pricing is structured so the foundations are never skipped to sell the shiny layer. If you're not sure which stage you're at, an AI visibility check will tell you.
Generally no. The disciplines share the same foundations, and splitting them invites contradictory work: one team optimising for rankings while the other restructures the same pages for citations. One programme with both goals is cheaper and more coherent.
Done properly, no. The core GEO moves (clearer structure, explicit facts, better third-party signals) are the same things Google rewards. Work that helps one and harms the other is usually a sign of gimmickry.
The terms overlap heavily. AEO usually refers to optimising for direct-answer surfaces, GEO for generative engines broadly. In practice the workstreams are the same; don't pay twice for two names.
You're better placed than most, but not automatically named. Ranking gets you retrieved; citations also depend on explicit facts, entity consistency and independent proof. Plenty of page-one businesses are absent from AI answers in their own category.