Writing for GEO feels like early SEO all over again

A client I’ve been creating online content for recently asked me to shift priorities. Instead of writing for traditional SEO first, they want me to focus on AEO first, with SEO as a secondary consideration.

I understand the logic. Discovery is changing. Content isn’t just competing for blue links anymore. It’s competing to be cited by AI systems.

Most of us are fluent in SEO by now. We’ve been living with it for decades. But the newer terms floating around — AEO, GEO and LLMO — are still settling, and they’re often used interchangeably. They’re related, but not identical:

• Search engine optimization (SEO): Optimizing content to rank in traditional search engine results pages.
• Answer engine optimization (AEO): Structuring content so it can be extracted and surfaced as a direct answer in AI-driven interfaces.
• Large language model optimization (LLMO): Optimizing content specifically for retrieval and citation by large language models.
• Generative engine optimization (GEO): A broader term encompassing optimization for AI systems that generate synthesized responses.

And that’s where the déjà vu kicked in. In 2004, I was leading a major B2B website rewrite. SEO was still new, and I was handed a document listing every keyword variation that had to appear a specific number of times, including every permutation of email: e-mail, email, Email, E-Mail, EMail.

I couldn’t do it. Alternating between e-mail in one paragraph and email in the next felt sloppy and mechanical. It looked like writing for a robot, not a reader.

Of course, SEO has evolved dramatically. It moved from keyword stuffing to intent, authority and quality signals. It’s matured. But right now? Writing for AEO feels a little like those early SEO days — it feels clunky and not natural. More structured, more explicit, more “just the facts.”

Here, I’ll share what’s different, what I’ve learned in practice and a side-by-side SEO vs. AEO example you can use. Because if we’re going to optimize for machines again, we should at least do it intentionally.

Traditional SEO vs. AEO: What’s actually different?

Let’s start with what hasn’t changed.

• We’re still trying to be found.
• We’re still trying to be useful.
• We’re still trying to demonstrate authority.

What’s changing is the mechanism of discovery.

Traditional SEO: Optimize for ranking

Traditional SEO is about earning a position on a search engine results page (SERP). That means targeting keywords and intent clusters, building authority and earning clicks.

Even as Google evolved from keyword matching to semantic search and intent modeling, the objective stayed the same: rank high enough that someone chooses to click. 

Your job as a writer was to balance clarity, persuasion and optimization — ideally without anyone noticing the optimization part.

AEO, LLMO and GEO: Optimize for extraction

AEO changes the target. Now you’re not just trying to rank. You’re trying to be used. Language models don’t present 10 blue links and ask users to choose. They synthesize, summarize and answer. Practically, that shifts how we write:

• Headings should mirror actual questions.
• The first sentence should answer directly.
• Sections should stand alone if excerpted.
• Lists often outperform narrative buildup.

Most AEO guidance I’ve seen is directionally proper — clear structure, direct answers, declarative language. But it can feel rigid. There’s less room for narrative setup or the slightly wry aside, which I happen to enjoy. That doesn’t mean the voice disappears. It means the priority order changes.

Traditional SEO asks: “How do we rank for this topic?” AEO asks: “If an AI system quoted three sentences from this section, would they stand alone as a complete, accurate answer?” That’s a different mental model that changes behavior.

Why this feels like early SEO and why that matters

If you’ve been in digital marketing long enough, you develop a kind of pattern recognition. AEO is triggering mine.

Back in 2004, when I was leading that B2B website rewrite, SEO guidance was mechanical. We weren’t talking about intent, authority or content quality. We were talking about density, frequency and exact match. The assumption was simple: search engines rewarded repetition.

But it didn’t take long for the industry to realize something important: optimizing purely for machines creates brittle content. It ranks until it doesn’t. It feels strategic until the algorithm evolves. Sound familiar?

Today, I’m seeing guidance that says:

• Every subhead should be a question.
• Every section should open with a direct answer.
• Paragraphs should be two to three sentences max.
• Avoid narrative buildup.
• Eliminate ambiguity.

But when advice becomes formula, formula becomes temptation. That’s where the déjà vu sets in. Early SEO encouraged us to write around keywords. Early AEO risks encouraging us to write around extraction patterns. The danger is over-optimization.

SEO matured dramatically. It moved from density targets to intent modeling, entity recognition, authority signals and user experience. It began rewarding depth, expertise and genuine usefulness. There’s no reason to believe AEO won’t follow a similar path.

But right now? We’re in the early “figure out the signals” phase. That means two things:

• There’s opportunity for those who adapt thoughtfully.
• There’s risk for those who overcorrect.

The goal isn’t to strip your writing of personality and nuance in pursuit of extractability, but to make your expertise easy to understand for both humans and machines. It’s a subtle distinction — and it matters.

What writing AEO-first looks like in practice

Once I moved from theory to actual client work, the shift became concrete. Writing AEO-first isn’t about sounding robotic. It’s about making expertise easier to extract without stripping away meaning. Here’s the framework I’m using now.

1. Start with the question

Subheads should mirror real user queries. Instead of clever phrasing or thematic transitions, think:

• What is answer engine optimization?
• How does AEO differ from SEO?
• Why does structure matter for LLMs?

Language models look for recognizable Q&A patterns. When your headings align with natural questions, you reduce friction for extraction. Clarity beats creativity, for now, at least at the structural level.

2. Answer immediately

The first one to two sentences under a subhead should directly answer the question — not set it up, not tease it. Answer it. For example:

• “Content is more likely to be cited by AI systems when it provides direct, clearly structured answers to specific questions.”

Then expand and add nuance. If your core point lives in paragraph three, you’re making it harder for a machine to use.

3. Make every section excerpt-ready

Ask yourself, “If an AI system quoted three sentences from this section, would they stand alone clearly and accurately?”

If not, tighten. Each section should function independently without requiring a surrounding narrative for context.

4. Be specific, not sweeping

Stronger:

• “In testing across 40+ campaigns…”
• “According to Google’s Search Quality Rater Guidelines…”
• “This typically fails when…”

Weaker:

• “Many experts believe…”
• “It’s often suggested…”

Language models are more likely to cite content that makes explicit, confident claims supported by detail. Vagueness is hard to summarize.

5. Anticipate the next question

If you define a term, follow with:

• Why it matters.
• When to apply it.
• Common mistakes.
• How it differs from adjacent concepts.

Language models reward completeness. A partial answer is less helpful than a connected one.

6. Avoid ‘keyword stuffing 2.0’

Here’s the caution. Early SEO over-optimized for density. Early AEO risks over-optimizing for format.

Warning signs:

• Every subhead feels forced.
• Every paragraph follows the same sentence rhythm.
• The writing sounds like it was generated to satisfy structure rather than express insight.

Structure is a tool. Formula is a trap.

7. Use AI as a stress test, not a substitute

Yes, I’m using AI in this process. But here’s the distinction: AI can help identify structural gaps, flag vagueness and suggest follow-up questions. It shouldn’t replace strategy, expertise or final judgment. That still belongs to the human in the room.

SEO vs. AEO: A side-by-side example

To make this concrete, let’s look at a short excerpt from one of my most-read MarTech articles: “3 marketing use cases for generative AI that aren’t copywriting.” 

Here’s a trimmed version of the original section, written with traditional SEO in mind.

Original version (SEO-oriented)

“Subhead: The value AI provides

“Many people talk about the value of generative AI in terms of the quantity of information it ingests. For me, it’s in its ability to quickly ingest and synthesize information. We’re directing it to specific, relevant information and asking it to digest it, identify what’s essential for the project and deliver the output in a usable format.

“For this use case, generative AI increased my productivity while maintaining the quality of the deliverable.”

Now here’s how I might rewrite that same idea for AEO-first optimization.

AEO-first version (structured for extraction)

“How does generative AI improve marketing productivity?

“Generative AI improves marketing productivity by quickly synthesizing targeted information into structured, usable outputs.

“Instead of manually reviewing multiple sources, marketers can direct AI to specific materials, ask it to extract what’s most relevant and return organized insights in minutes.

“In campaign planning and foundational analysis work, this reduces production time while maintaining output quality.”

What changed and why

The substance didn’t change. The structure did. 

• The subhead became a question. 
• The first sentence became a direct answer. 
• The explanation tightened into standalone, excerpt-ready paragraphs. 

The AEO version is more explicit and easier to extract. That’s the trade-off. You gain citation potential. You risk sounding like everyone else.

The bigger question: What happens to voice?

If you’ve been in marketing long enough, you’ve seen this cycle before.

A new discovery mechanism emerges. We reverse-engineer it and optimize hard. Then the system evolves to reward quality again. SEO did.

It moved from density targets and awkward keyword permutations to intent, authority and user experience. It matured, and the content that endured wasn’t the most mechanically optimized. It was the most genuinely useful.

I expect AEO will follow a similar arc. Right now, we’re in the early phase. It’s also where over-correction tends to happen. You gain citation potential and risk sounding like everyone else.

In marketing, sameness is rarely a competitive advantage. The goal isn’t to strip your writing of personality. It’s to make your expertise easier for humans and machines to understand.

Optimize for answer engines, structure thoughtfully and pressure-test for extractability. But don’t outsource your thinking and don’t flatten your perspective. The objective isn’t to sound like a language model. It’s to be the best answer.

AI disclosure paragraph

I used generative AI to help structure portions of this article, to edit it for length, and to pressure-test its AEO readiness (full disclosure: it’s AEO-aware, not fully AEO-optimized, which is appropriate given my take). The experiences, strategy, perspective and conclusions are mine. Given that I’ve written previously about responsible AI disclosure, it felt appropriate to say that plainly here.

The post Writing for GEO feels like early SEO all over again appeared first on MarTech.