In early 2023, I started noticing something unusual in our analytics. Certain pages were holding their traffic — even growing — while our traditional rank-tracking tools showed flat or declining keyword positions. At the same time, I was testing ChatGPT Search and Perplexity and noticing that our product content was being referenced in AI-generated answers for queries we’d never specifically targeted.

It wasn’t a coincidence. It was the result of a content strategy I’d been building for months, one that I’ve since formalised into what I call our GEO/AEO implementation framework. This post is that framework — the actual playbook, not the theory.

If you’ve already read my introduction to GEO and AEO, this is the practical follow-up. If you haven’t, the short version is this: optimising for AI-driven search (GEO) and answer engines (AEO) is increasingly important as tools like Google AI Overviews, ChatGPT Search, and Perplexity change how people find and consume information.

Let’s get into the specifics.

Step 1: Build your “Answer Architecture” before you write a word

The biggest mistake I see content teams make is treating GEO/AEO as something you bolt on after writing. It doesn’t work that way. You need to plan your content around answers before you start.

Here’s the process I use.

Map questions, not just keywords. Traditional keyword research gives you search volumes and difficulty scores. That’s still useful, but for AEO it’s not enough. You need to understand the specific questions behind the keywords. A keyword like “VPN for Windows” is a topic. “Is a VPN legal in India?” is a question. AI systems are much more likely to cite content that directly addresses a specific question than content that generally covers a topic area.

I use three sources for question mapping:

  • Google’s “People Also Ask” boxes for my target keywords
  • AnswerThePublic for long-tail question variations
  • Reddit and Quora threads in my niche — the questions people ask in conversational forums are often better predictors of AI query patterns than keyword tools

Create a question matrix. Before writing any piece of content, I list every question that piece needs to answer — typically 8 to 15 questions per article. Each section of the article then maps to one or more of those questions. If a section doesn’t answer a specific question from the matrix, I either cut it or rework it until it does.

This sounds rigid, but it produces better content. It forces you to be useful rather than just comprehensive.

Step 2: Structure your content for AI extraction

AI systems don’t read content the way humans do. They parse it. They look for signals that indicate what a piece of content is about, who wrote it, and whether specific claims are backed up. Here’s how to structure your content to make that extraction as clean as possible.

Use descriptive H2 and H3 headings that are themselves answers. Instead of a heading like “Our Approach”, write “How to build a GEO-optimised content strategy in 5 steps”. The heading should tell the AI exactly what the section contains. This is also good for featured snippets — Google often pulls a heading plus the content immediately below it.

Write a direct answer first, then explain. In every section, lead with the direct answer, then provide context and elaboration. This is the opposite of how many writers are trained — building up to a conclusion. For AI-optimised content, the conclusion comes first.

For example:

Weak for AEO: “There are many factors that affect how quickly you’ll see results from an SEO campaign. These include the age of your domain, the competitiveness of your industry, your content quality, your technical SEO health, and your link profile. Taking all of these into account…”

Strong for AEO: “Most SEO campaigns show measurable results within 3–6 months for established sites and 6–12 months for new domains. The main variables are domain authority, keyword competition, and how consistently you publish high-quality content. Here’s what affects each…”

Notice that the second version gives a specific, usable answer in the first sentence. An AI can extract that sentence and use it directly. The first version requires the AI to synthesise across several sentences to form a coherent answer — which it might do incorrectly or not at all.

Use tables for comparative information. When you’re comparing options, tools, or approaches, put that information in a table. Tables are extremely effective for AI extraction because they create explicit relationships between data points. A table comparing five VPN providers across five criteria is far more citable than five paragraphs making the same comparisons in prose.

Add FAQ sections at the end of every post. This is non-negotiable in my content workflow. Every post gets a FAQ section with 5–10 questions that:

  • Include questions the article answers but doesn’t explicitly frame as questions
  • Include adjacent questions the reader might have after finishing the article
  • Are written in the exact natural language someone would type or speak into a search query

These FAQ sections are then marked up with JSON-LD schema.

Step 3: Implement the technical schema stack

Schema markup is the infrastructure layer of AEO. It tells search engines — and increasingly, AI systems — the structured facts about your content. Here’s the schema stack I implement for every content-focused website.

Article schema on every blog post. At minimum, include: headline, author (with a reference to your Person schema), datePublished, dateModified, and publisher. The author link is important — it connects the article to the entity that wrote it, which builds author authority over time.

Person schema on your About page and author page. This is where you define the entity of the author — their name, their expertise, their social profiles, their affiliation. Think of it as telling Google’s knowledge graph who you are. Include sameAs links to your LinkedIn, Twitter, and any other authoritative profiles. This is how AI systems learn to associate you with specific topics.

FAQPage schema on every post with a FAQ section. Use JSON-LD format. If you’re on WordPress, Rank Math handles this natively. If you’re hand-coding it, the structure is straightforward — an array of Question and Answer pairs.

BreadcrumbList schema on every page. This helps AI systems understand your site’s content hierarchy and the relationships between topics.

Organisation schema on your homepage. Define your brand as an entity — name, URL, logo, contact information, and sameAs links to your social profiles and any business directories.

One practical note: use Google’s Rich Results Test and Schema Markup Validator to verify your implementation. Broken schema is worse than no schema — it can create confusion about what your content actually is.

Step 4: Build citation-worthy content assets

Getting cited in AI-generated answers isn’t just about structure — it’s about having content that genuinely deserves to be cited. Here’s what I’ve found works.

Original data and first-hand experience. This is the single highest-leverage investment you can make. AI systems are trained to identify and prefer primary sources. If you have data that doesn’t exist anywhere else — your own survey results, your own campaign metrics, your own test results — publish it. I’ve had posts with relatively low traditional SEO metrics perform extremely well in AI citations simply because they contained original industry data that the AI couldn’t find anywhere else.

Clear attribution and credentials. State who you are and why you’re qualified to speak on the topic. Not in a self-promotional way, but in a factual, direct way. “Based on managing $50,000+ per month in Google Ads across consumer software verticals, here’s what I’ve found about bid optimisation…” This kind of attribution signals to an AI that this content comes from a practitioner with actual experience, not someone summarising what they read elsewhere.

Comprehensive coverage of a topic. AI systems are more likely to cite a resource that covers a topic thoroughly than one that covers it briefly. This doesn’t mean padding — it means genuinely addressing all the sub-questions around a topic in one place. Comprehensive resources become reference documents that AI systems return to repeatedly.

Regular updates with a datestamp. AI systems increasingly prefer fresh information. Add a “Last updated” date to your posts and actually update them when the information changes. A post about GEO/AEO published in 2024 and updated in 2025 with new examples is more citable than one that’s clearly two years stale.

Step 5: Monitor your AI visibility

Traditional SEO monitoring — rank tracking, traffic analytics, backlink profiles — doesn’t capture your AI visibility. You need additional monitoring.

Test your queries manually and regularly. Every week, I run a set of test queries through ChatGPT, Perplexity, and Google (with AI Overviews enabled) and look at whether our content is being cited or whether our brand is being mentioned. This is manual work, but it’s the most direct signal available right now.

Track branded mentions in AI outputs. This is harder to systematise, but I flag any time a prospect or customer mentions they “heard about us” from an AI recommendation. This qualitative signal, combined with the direct testing above, gives a reasonably clear picture of your AI visibility.

Watch your featured snippet wins. Featured snippet performance is a reliable proxy for AEO health. If you’re winning featured snippets in your niche, you’re producing the kind of structured, direct content that AI systems also prefer. Google Search Console shows your featured snippet impressions and clicks under “Search Appearance”.

Monitor your People Also Ask appearance. Tools like SEMrush and Ahrefs now track PAA box appearances. Being a consistent source of PAA answers strongly correlates with AI citation, in my experience.

The content types that perform best

After two years of implementing this across multiple product categories and content types, here’s my ranking of what earns AI citations most reliably.

Definitive guides with original perspective perform best. These are long-form (2,000+ words), comprehensive, and include first-hand insight that isn’t available elsewhere. They take longer to produce but have the longest citation lifespan.

Comparison articles perform very well, especially when they include a clear recommendation rather than sitting on the fence. AI systems are often trying to help someone make a decision — they prefer sources that actually make the call.

Data-backed how-to content performs consistently. Step-by-step guides with specific, actionable instructions and clear examples are extremely citable. The key is specificity — “increase your click-through rate” is less citable than “increase your click-through rate by testing three subject line formats over 30 days”.

Glossary and definition content is underrated. When an AI needs to explain what a term means, it looks for clear, authoritative definitions. A well-structured glossary post on your site — defining 20–30 terms in your niche — is a surprisingly effective way to build AI citation volume.

Opinion and prediction content is hit or miss. Specific, well-reasoned predictions (“Here’s why GEO will become standard practice for SaaS brands by 2026, and what it means for content strategy”) can perform well. Vague thought leadership doesn’t.

A note on what doesn’t work

A few approaches I’ve tested that didn’t deliver meaningful results:

Keyword stuffing in FAQ sections. The algorithm is smarter than that. Writing FAQ content that sounds natural and genuinely addresses the question matters far more than keyword density.

Duplicate FAQ content across multiple pages. Having the same FAQ schema on ten pages targeting the same questions doesn’t multiply your citations — it dilutes your authority. One definitive answer per question, on the most relevant page.

Chasing AI citations at the expense of human readability. The best GEO/AEO content is also the best content for human readers. These aren’t in conflict. If you’re making your writing less readable in order to score points with an AI, you’re optimising for the wrong thing.

Where to start if you’re doing this from scratch

If you’re reading this and thinking “we’ve done none of this”, don’t panic. Start with the two things that have the highest impact for the least effort:

First, go through your five highest-traffic pages and add a FAQ section to each one, marked up with FAQ schema. This alone can generate featured snippet wins and AI citations within weeks.

Second, identify one topic in your niche where you have genuine first-hand expertise and write one comprehensive guide — not for SEO, but for the person who genuinely wants to understand that topic. Make it the best resource on that topic that exists. Publish it, date it, update it quarterly.

Everything else builds from there.

Need help implementing GEO/AEO for your business? Whether you’re starting from scratch or auditing an existing content strategy, I offer hands-on consulting to get your brand visible in Google AI Overviews, ChatGPT Search, and Perplexity. Let’s connect and build your AI search playbook together.