Since search engines started answering with AI, the field has filled up with acronyms: GEO, AEO, GAIO, LLMO, AISEO. Each new one revives the same argument. Part of the SEO community says we are facing a brand-new discipline that replaces SEO; another part sees recycled marketing to sell courses under a fresh label.
My take is that GEO is not something new from scratch: it is SEO specialized for the language-model era. The foundations hold. What changes is the focus, plus two layers that did not weigh as much before, brand and measurement.
The debate
GEO (Generative Engine Optimization) means optimizing so that a generative engine uses and cites you in its answer. The term comes from a 2024 academic paper; AEO is older. Behind the acronyms sit two camps.
On one side, those who hold that this is still SEO. Natzir Turrado titled his piece bluntly, “It’s not GEO or AEO, it’s just SEO,” and argues that optimizing for large language models is, in essence, the logical evolution of classic SEO. His historical reminder is hard to argue with: nobody renamed SEO for mobile-first indexing, for featured snippets, or for structured data.
Carlos Ortega is blunter still: it all sounds a lot like SEO to him, and he insists that to show up in AI results what you need is good content and the same old reliable SEO.
Lily Ray, at Amsive, frames it with more nuance. For her, AI search is “a classic evolution of SEO,” on par with voice search or mobile-first, and she warns about the “GEO grifters” who repackage core SEO approaches under a different name. But she adds a piece I find key: GEO is not abandoning SEO, it is “a new system for competing for, capturing, and measuring success across AI platforms.”
On the other side stands the view that this really is something new, argued seriously by Mike King of iPullRank, who has stopped calling it SEO: he now offers “Relevance Engineering.” His distinction is sharp: SEO was about ranking on the SERP; GEO is about “being part of the answer, whether or not a link is shown.” He sums it up for the reluctant: “the SEO industry is being pulled reluctantly into the GEO era” (via Technology Magazine).
And here is what almost nobody highlights: even King admits “the fundamentals still need to be here” (in a Search Engine Land interview). He does not say SEO is useless. He says what the engines do with those fundamentals is radically different, and that this already justifies a new discipline. That is where I get off: doing something different with the same foundations is specialization, not a separate discipline. It is exactly Natzir’s mobile-first point: changing tactics on the same base never forced anyone to rename SEO.
My take: GEO is SEO, with the focus somewhere else
I land in the middle. I agree with Natzir and Carlos on the essentials: the foundation is SEO, and anyone selling GEO as a clean break that forces you to throw out everything you know usually has a course to sell you. But I will not reduce it to “just SEO,” the way Carlos does: those two layers, brand and measurement, are real and they change my day-to-day work. Neither “nothing has changed” nor “everything is new.”
I see it every week, which is why I teach it this way in my classes at BIG School. In real audits I have found the same content gets picked up or ignored depending on how extractable it is: the passages a model cites tend to answer the question in under fifty words and do not depend on the previous paragraph. I have watched GPTBot and ClaudeBot pull the HTML in my logs but miss whatever JavaScript injects. And I have had to build my own tools, in Python on top of the APIs, to measure something no dashboard hands you: whether a specific person, asking a specific model, sees you or not.
That last part, measurement, is not my invention. Lily Ray names it without meaning to when she defines GEO as a new system for measuring success. The brand layer, by contrast, I defend as my own thesis: in a setting where the model breaks ties by entity recognition, your brand matters more than ever.
What actually changes when you optimize for AI
When I bring the thesis down to practice, what changes sorts into four pillars. I teach them in this order:
- Solid SEO is the foundation. There is no separate “AI index”: for a model to cite you, you first have to be in Google’s index and a valid candidate. Ranking does not guarantee the citation, but it is the entry condition.
- Brand amplifies. Once you are a candidate, what breaks the tie is mentions and recognition of your entity. In my audits, content with a recognized brand behind it gets picked up where the same text, without that brand, did not; that is why I work on my own entity before anything else. I put this as a working hypothesis, not a law: it is the leg I still have to close with public data.
- Extractability decides. The model has to be able to read your page and lift a citable fragment from it. This is where the technical work concentrates: that the content does not depend on JavaScript the AI crawlers still do not render, that passages are self-contained, and that semantic relevance is measurable. When the content lives in documents, preprocessing matters as much as the HTML, and all of it connects to how models retrieve and use your content, with and without RAG.
- Measurement is per person. You do not have “a ranking in ChatGPT”: you have results per engine, per segment, per query. Most of what decides the citation lives inside the model, out of your control, so the only way out is to measure directly, not infer it from organic rankings.
Two of those pillars, foundation and extractability, are plain old SEO done with more finesse. The other two, brand and measurement, are the new layers. That is all the novelty, and it is not trivial, but it is not a separate discipline.
What does not change (and why it matters)
Here is a continuity check I like: according to Peter Raventós, around 80% of what you do to rank in Google also works in AI. So the most expensive mistake I see is the opposite of the one people fear: people who neglect their technical SEO and their authority while chasing “GEO” tricks, and end up with neither. If the foundations are weak, no new acronym will hold them up.
Further reading
FAQ
Is GEO a new discipline or is it SEO?
In my view, it is not a new discipline. It is SEO specialized for AI search engines and language models: the foundations are the same and what changes is the focus on how they read and cite you.
How is GEO different from SEO?
At the foundations, not at all: GEO rests on the same technical, content and authority SEO. The difference is two layers that classic SEO did not prioritize as much: brand (the model breaks ties by entity recognition) and measurement (there is no single ranking; you measure per engine and per person). That is why I argue GEO is specialized SEO, not a separate discipline.
How is GEO different from AEO?
The acronyms first: AEO is Answer Engine Optimization and GEO is Generative Engine Optimization. AEO is the older, broader idea: optimizing to be the direct answer, including non-generative formats like featured snippets or voice search. GEO is more specific: optimizing so a generative engine built on an LLM, such as AI Overviews, ChatGPT or AI Mode, uses and cites you. In practice they converge.
Does classic SEO still help for AI?
It is the base. Search intent, quality, architecture, authority and structured data are still what holds up any strategy, including the AI one.
How do I get ChatGPT or AI Mode to cite me?
There is no shortcut, but there is a method: be in Google’s index, have a recognized entity, and write passages a model can extract and cite without ambiguity. The detail of what to prioritize, what to measure and with which tools is what I work through across the rest of the blog. If you care about judgment and strategy, not just the tactic of the week, follow along.
