An Honest Take Before We Start
I spent a good chunk of 2021 and 2022 selling clients on the pillar-cluster model. Broad pillar page at the centre, 15 cluster articles pointing in and out, internal links woven through the whole thing. It worked. Traffic climbed, rankings stabilised, and the logic was clean enough to explain to a CFO in four minutes.
So when I say the model is dead, I'm not being contrarian. I built my own site on it. I'm saying that the mechanism which made it work — users clicking through ten blue links to navigate a search result page — has been structurally disrupted. AI Overviews now appear in over 20–30% of Google searches, and that number moves up every quarter. When the answer appears above the results, the click-through journey changes. Your content architecture has to change with it.
"The pillar-cluster model optimised for the journey users took through Google. AI Overviews shortened that journey. The model needs to catch up."
This is not a "SEO is dead" argument. Organic search still drives more qualified traffic than almost any other channel for most businesses I work with. But the rules of the game have shifted, and the playbook from 2021 will not get you where you want to go in 2026. Let's get into what actually does.
What Actually Broke — and Why
The original topic cluster model, pioneered at scale by HubSpot around 2017, was built on a specific assumption: that search engines reward topical depth, and that a hub-and-spoke architecture signals that depth better than isolated keyword pages. Both of those things are still true. The problem is what "depth" means to a search engine has changed significantly.
In 2021, depth meant: does your site have comprehensive coverage of a topic, connected through internal links? In 2026, depth means: can an AI system extract, validate, and confidently cite your content across a range of related queries without going elsewhere?
That's a different question. And it requires a different architecture to answer it.
The thing that trips people up: clusters still matter. Interconnected content still beats isolated keyword pages. What changed is why they matter and how they need to be structured. You're not building a hub for humans to navigate anymore. You're building a knowledge graph for AI systems to read, validate, and cite.
The Query Fan-Out Problem
Here's the mechanism that makes this concrete. When a user types a query into Google in 2026, Google's AI doesn't run one search. It performs what researchers call query fan-out — it expands the original query into 12 to 15 related sub-queries in the background, simultaneously. It's looking for the "how," the "why," the "cost," the "comparison," and the "what if" all at once.
Your old pillar page was built to answer one of those sub-queries well. The new requirement is that your content architecture answers enough of them — credibly, consistently, with sufficient depth — that the AI system treats your domain as the default reference for the topic.
Read those numbers again. Especially the last one. Being cited in AI Overviews isn't about having the most comprehensive single page on a topic. It's about your site demonstrating that it covers the topic's ecosystem — the full territory of related questions someone might ask.
What Replaces the Old Model
Four shifts define the new architecture. None of them are radical departures from good SEO thinking — but each one requires updating how you approach content planning, structure, and production.
Topical Authority Is Not New — But It Works Differently Now
Topical authority as a concept has been around for years. What changed after Google's March 2026 Core Update is how it's measured. A site with 20 interconnected, regularly updated articles on a specific topic now consistently outranks a site with a single comprehensive guide on the same topic — even when the single article is better written.
That's worth pausing on. Better writing loses to better architecture. Because AI systems don't evaluate pages in isolation. They evaluate your site's entire topical footprint — the breadth and depth of your coverage — when deciding whether to cite you.
- Pick your most important topic. Count how many pages on your site address a genuine sub-question within it.
- If the answer is fewer than five, you don't have topical authority — you have a pillar page and good intentions.
- Run each of those pages through Google's AI Overview for its target question. Are you cited? If not, the page lacks extractability — it answers the question, but not in a way an AI can cleanly pull from.
- Check internal linking: does every cluster page link back to the hub, and to at least one related cluster page? If no, your architecture is still one-directional.
The businesses seeing the sharpest organic growth right now — in verticals I work in directly, including healthcare, professional services, and e-commerce — are the ones that treated topical authority as infrastructure, not output. They built the architecture before they built the content.
GEO and AEO: The New Acronyms That Actually Matter
Two terms have entered the SEO vocabulary in 2025–2026 that describe the same underlying shift from different angles.
GEO (Generative Engine Optimization) is the practice of optimising content to appear in AI-generated answers — ChatGPT, Gemini, Perplexity, Google AI Overviews. Unlike traditional SEO, where ranking #1 guarantees clicks, GEO requires your content to be structured so AI systems can extract, paraphrase, and cite it accurately.
AEO (Answer Engine Optimization) focuses on the question-answer structure of content. The core premise: if you can't identify the specific question each section of your content answers, an AI system probably can't either — and won't cite it.
SEO gets you on the results page. GEO gets you cited in the AI Overview above it. AEO determines whether your content can be extracted clearly enough to be quoted accurately. You need all three — but AEO is where most content currently fails. If your answer is buried under 300 words of setup, the AI will pull from someone else's cleaner version.
The New Architecture: What to Actually Build
Here is the practical version — the content architecture I'm now building for clients instead of the old pillar-cluster model. Call it the topical knowledge network.
Map the full question territory before writing anything
Start with your core topic and enumerate every genuine question a user might ask across the full journey — discovery, consideration, decision, and post-purchase (or post-conversion, for services). You're mapping a territory, not a keyword list. Group questions by intent cluster, not search volume. This becomes your content architecture.
Build your authority hub — not a pillar page
The hub is not 3,500 words that answer everything. It's 1,200–1,800 words that define the topic clearly at the entity level, establish your position on it, and explicitly routes to each cluster page. Think of it as an expert's introduction that tells you exactly where to go for depth — not a document that tries to provide all the depth itself.
Write cluster pages that own one question completely
Each cluster page answers one specific question — and answers it so thoroughly that an AI system can extract a clean, accurate answer from it without reading the rest of your site. Use structured headings, direct answers in the first 150 words, and FAQ schema. These pages should feel like the best single answer on the internet for their specific question, not like a chapter in a longer document.
Wire the network bi-directionally
Every cluster page links back to the hub and to at least two related cluster pages. The hub links to every cluster. Add entity-level context to your internal anchor text — not just "click here" or the page title, but a phrase that explains why the related page is relevant from the current context. This is the difference between a website and a knowledge graph.
Add structured data to every page in the network
At minimum: Article schema on the hub, FAQ schema on every cluster page, and BreadcrumbList connecting the network. For healthcare, legal, or financial topics, add appropriate specialist schemas. This is how you speak directly to AI crawlers — JSON-LD is the closest thing to a language AI systems already know. If the schema is missing, you're relying on the AI to infer structure that you could have made explicit.
Treat freshness as infrastructure, not a publishing schedule
Stale data costs you AI citations. An article citing 2023 statistics in May 2026 will be passed over in favour of one that has been updated. Build content update cycles into your production process the same way you build publishing cycles. Track your AI citation rate separately from your keyword rankings — they are different signals and require different interventions.
Using Claude to Audit Your Existing Content Architecture
If you've been running the old pillar-cluster model, you don't necessarily need to tear it down. Most sites I audit have solid content that just isn't structured for AI extractability. The work is often architectural — rewiring internal links, adding schema, restructuring answers to lead with the direct response — rather than starting from scratch.
This is where Claude is genuinely useful. Feed it your existing content structure and it will identify which pages have extractability gaps, where your internal linking is one-directional, and which cluster topics you're missing entirely.
Export your top 20–30 pages by organic traffic from GA4 or Search Console. Include: page URL, title, primary keyword, word count, internal links in/out (if available). The more structured your input, the more specific Claude's output. This is not a magic wand — it's a fast-cycle analysis layer. Your domain expertise still determines what recommendations actually apply to your business.
The Claude AI Prompt: Audit Your Content Architecture for the AI Era
What the Old Model Got Right — and What to Keep
I want to be fair here, because some of what you built under the old model is genuinely worth keeping.
The core insight — that connected content beats isolated keyword pages — is more true in 2026 than it's ever been. The data on AI citations is unambiguous: 86% of AI citations came from sites with five or more interconnected pages on a topic. Clusters work. They just need to be rebuilt around extractability and entity coherence rather than hub-and-spoke navigation.
What to audit and keep: any cluster page that answers a single specific question with depth and a clear direct answer in the opening paragraph. These are already structured for AI extractability — they just need schema added and internal links tightened.
What to rethink: any content built around a broad head keyword rather than a specific question. These pages typically have high word counts, decent rankings for vanity metrics, and almost no AI citation rate. They're doing less work than they look like they're doing.
EEAT, AEO, GEO — and What Akif Actually Checks for in Audits
The things I actually check first in a content audit in 2026: (1) Does the answer to the target question appear in the first 150 words? (2) Is FAQ schema present? (3) Are internal links bi-directional? (4) When was this last updated, and does the content reflect current data? (5) How many pages on this domain address related sub-questions?
Keyword density, meta description length, exact-match anchor text — these get maybe 10% of my attention now. The other 90% is architecture and extractability. That's the shift.
The Bottom Line
The topic cluster model isn't dead in the sense that everything you built is worthless. It's dead in the sense that the version of it from 2021 — broad pillar pages, one-directional internal links, keyword-led briefs — will not get you AI citations, and AI citations are becoming a primary organic traffic driver.
The replacement is more rigorous, not less. You need tighter question mapping, more specific cluster pages, cleaner internal linking, structured data on everything, and a freshness process that keeps your data current. That's more work, not less. But the sites doing this work are seeing 40–300% organic traffic growth in verticals I track — and more importantly, they're building the kind of authority that compounds over time rather than eroding as AI search expands.
Build the architecture first. Then build the content. That sequence matters more than it used to.
.webp)
Driven by advanced SEO expertise, deep marketing analytics, high-impact content strategy
With 5+ years of hands-on experience, I specialize in holistic search strategies that don’t just rank—they drive real, measurable business growth. I’ve worked across industries including healthcare, hospitality, legal, e-commerce, and professional services, helping brands dominate their target markets. My approach bridges the gap between raw data and creative execution. Every strategy I build is rooted in rigorous market analysis, structured SEO frameworks, and tailored content ecosystems—no templates, no shortcuts. Whether you’re a single-location brand or scaling across multiple cities, I create data-driven marketing systems designed to compound results and grow with you.
Ready to rebuild your content architecture?
Copy the Claude audit prompt above, paste your page inventory, and share this guide with any content strategist still building pillar pages the 2021 way.
Jump back to the Toolkit →