How to Restructure a Blog for AI Search
Key Takeaways
Most B2B blogs aren't invisible to AI because the writing is bad — they're invisible because the architecture is broken. No pillars, no clusters, no signals that tell AI what the site is actually an authority on.
73% of B2B websites saw significant organic traffic loss between 2024 and 2025. Unstructured blogs are the easiest content for Google and AI Overviews to devalue.
Pillar pages give AI a single, comprehensive hub to map topical authority. Without one, even strong cluster posts read as isolated and unconnected.
Category and tag cleanup is a semantic signal. Meaningless categories like 'Blog' and 'Resources' tell AI systems nothing about your expertise.
Internal linking is the architecture layer that AI systems crawl. Every cluster post should link up to its pillar and sideways to 2–3 sibling posts.
FAQ blocks with schema markup are the single fastest GEO improvement you can make to any existing post — they're the blocks AI systems pull most often for direct answers.
Most existing B2B blogs (honestly, most blogs in general) are invisible to AI. They’ve invested in SEO for years, but we live in the era of GEO.
The content isn’t always bad. But they're invisible because the architecture is broken. Most blogs are written because someone said, “websites with lots of blogs are SEO optimized!”
But they’re posted with scattered topics. There are no clear pillars and no internal linking pattern.
These things tell AI systems what the site is an authority on. In 90 minutes, you can't fix everything, but you can redesign the blog’s structure so that AI and Google AI Overviews see coherent expertise instead of years of random publishing decisions.
Good SEO content and citable AI content are two different standards.
You can rank #1 on Google and be completely invisible to ChatGPT, Perplexity, and Google AI Overviews — because AI systems don't evaluate pages the way Google does. They evaluate content networks.
They look for semantic clusters, internal linking patterns, and comprehensive topic coverage. This means that a blog with 80 posts and no pillar structure looks like noise to an AI — even if each individual post is excellent.
Let me show you what I mean. Below, I'll walk through a 90-minute working session using a fictional but realistic B2B SaaS blog as the example — the kind of site I see and work on on a weekly basis.
What does a traditional B2B blog look like?
A typical broken B2B blog has 40–150 posts published over several years, with no consistent topical structure, no pillar pages, and category/tag systems built post by post rather than planned as part of a content architecture. The posts themselves may be well-written — the problem is structural, not editorial.
For this teardown, I'm working with a fictional company — let's call them Trackably, a B2B SaaS analytics platform. (Sounds real, right? Maybe it is. Sorry ahead of time!)
Here's what their blog looks like before we touch anything:
The 'before' picture:
~80 posts published over 4 years
Topics include: product analytics how-tos, customer onboarding guides, revenue forecasting tips, AI feature announcements, company news, and a few thought leadership op-eds
Navigation: Blog → one endless reverse-chronological feed
Categories: 'Blog,' 'Resources,' 'Insights' — no semantic difference between them
Tags: 50+ one-off tags with 1–2 posts each (e.g., 'Q3 2023,' 'webinar recap,' 'announcement')
No pillar pages. No hub-and-spoke structure. No explicit internal linking strategy.
To an AI system, this blog looks like a site with scattered opinions — not an authority. There's no signal that Trackably owns any particular topic deeply enough to be cited.
That's fixable. Here's how.
Want your own GEO architecture teardown?
I do this for B2B content teams who want their site to function as an AI-ready content library, not a blog graveyard.
Let's talk →Minutes 0–15
How do you decide what topics to own?
The first 15 minutes aren't about writing — they're about deciding. Export all existing posts, group them by broad theme, and identify 3–4 topics where you already have enough content depth to establish topical authority. These become your pillar candidates.
Before you touch a single post, you need a fast inventory of your existing blogs.
Open a spreadsheet (or export from your CMS) with: post title, URL, publish date, category, and tags. You don’t have to read every post — just skimming titles.
You can also do this by having an AI like Perplexity run a quick search of your site and create a list for you.
Step 1: Group by theme
Scan the titles and sort them into broad buckets. For Trackably, those buckets shake out like this:
Customer onboarding — 14 posts
Usage analytics & product insights — 19 posts
Revenue forecasting & reporting — 11 posts
AI features & product updates — 9 posts
Company news & miscellaneous — 12 posts
Other / off-topic — 15 posts
Step 2: Identify real pillar candidates
A theme qualifies as a pillar candidate if it has at least 6 substantial posts already written AND it maps to something the company actually wants to be known for (i.e., it connects to their product, their ICP, or both).
For Trackably, three pillars are clear:
Customer Onboarding Strategy
Usage Analytics & Product Insights
Revenue Forecasting & Reporting
The AI features content is growing, but not deep enough yet — it's a candidate for a fourth pillar to develop over time, not to restructure now.
GEO rationale: AI systems cluster your content embeddings the same way you just did — by semantic proximity. The difference is that you're now doing it intentionally instead of leaving it to chance. When AI detects a clear cluster around a topic, it's more likely to attribute expertise to that domain and cite your content in relevant answers.
Minutes 15–45
How do you build the pillar–cluster map?
The pillar–cluster map assigns each pillar a hub page and maps each existing post to its cluster. For each cluster, you identify whether the pillar page exists or needs to be created, and you note which posts cover the standard cluster content types: foundational, implementation, tools/templates, comparison, and ROI/metrics.
For each of the three pillars, I'm doing three things: designating or creating the pillar page, mapping existing posts into clusters, and flagging what's missing.
Here's what it looks like for the Customer Onboarding Strategy pillar:
| Content Type | Existing Post | Action |
|---|---|---|
|
Pillar page
What is / Overview
|
No dedicated pillar page exists | + Create new |
|
Foundational
Why it matters
|
"Why Onboarding Determines 90-Day Retention" | Keep + link |
|
How-to
Implementation
|
"5-Step Onboarding Checklist for SaaS Teams" | Keep + link |
|
Tools / Templates
|
"Onboarding Email Sequence Template" | Keep + link |
|
Comparison
X vs Y
|
Missing | + Create new post |
|
Metrics / ROI
|
"Time-to-Value Benchmarks for B2B SaaS" | Keep + link |
Run this exercise for each of the three pillars. At the end, you have:
A list of pillar pages to create (or upgrade from existing strong posts)
A cluster map showing where each existing post belongs
A gap list of content types that are missing per pillar
What about posts that don't fit any pillar?
For Trackably, 27 posts don't map cleanly to any priority pillar. Those get three options:
noindex if they're old and low-quality
consolidate if several thin posts cover the same ground and could become one strong piece
move to a news/updates lane if they're announcements that don't need organic visibility
This step is uncomfortable for content teams who've invested in those posts — but a blog with 50 well-clustered posts outperforms a blog with 80 scattered ones every time, for both SEO and GEO.
Minutes 45–75
How do you rebuild navigation, categories, and internal links?
The architecture you designed in the pillar map now needs to be expressed in the site's actual structure — in ways both humans and AI crawlers can see. That means pillar-aligned category URLs, a cleaned-up tag system, and an internal-linking pass that connects every cluster post to its pillar and sibling posts.
Rework categories into pillar lanes
Each pillar becomes a category with a clean, descriptive URL:
/blog/customer-onboarding-strategy/
/blog/usage-analytics/
/blog/revenue-forecasting/
Remove 'Blog,' 'Resources,' and 'Insights' as categories for evergreen content. They add zero semantic signal. A post categorized as 'Insights' tells AI nothing.
A post categorized as 'customer-onboarding-strategy' tells AI exactly what topical cluster it belongs to.
Clean up tags
We’re aiming for 10–20 tags that mirror real subtopics within your pillars. That means we’ll kill one-off and date-based tags.
For Trackably, that means deleting tags like '2023,' 'webinar recap,' and 'announcement' — and standardizing tags like 'time-to-value,' 'onboarding metrics,' and 'SaaS retention.'
Internal linking pass
This is the highest-leverage thing you can do in this window. For every post now mapped to a cluster:
Add a link to its pillar page at the top or bottom of the post with descriptive anchor text (e.g., 'our full guide to customer onboarding strategy,' not 'click here')
Add links to 1–2 sibling cluster posts that answer the next-step question
Add the phrase 'Part of our [Pillar Name] series' as a contextual callout near the top
Example: 'In our full customer onboarding strategy guide, we break down how to design the entire post-sale journey end to end.' → Linked to the pillar page.
Before vs. After: What the architecture change looks like:
| Element | Before | After |
|---|---|---|
| Blog navigation | Endless reverse-chronological feed | Pillar-aligned hub URLs (/customer-onboarding-strategy/) |
| Categories | 'Blog,' 'Resources,' 'Insights' | 3 pillar categories + 1 news lane |
| Tags | 50+ one-off tags, 1–2 posts each | 15 standardized topical tags |
| Internal links | Inconsistent, often none | Every cluster post → pillar + 2 siblings |
| Orphan posts | ~15 posts with no inbound links | Consolidated, noindexed, or linked in |
| AI visibility | Scattered, no clear authority signals | Coherent cluster structure, citeable network |
Minutes 75–90
What GEO enhancements make the biggest difference, the fastest?
The final 15 minutes are for GEO-specific enhancements that improve AI citability without requiring a full rewrite. The three highest-impact moves: add FAQ blocks to your pillar pages, add author metadata and last-updated dates to your top posts, and add Article and FAQPage schema to every pillar and cluster page.
You're not rewriting content in this window — you're adding the structural signals that help AI systems extract and cite what's already there.
Add a short FAQ block to each pillar page and your top 5 cluster posts. 4–6 questions, answer-first, plain language. These are the blocks AI systems pull most often for direct answers.
Add last-updated dates, author bios, and a one-line 'who this is for' statement to every pillar page. Author metadata linked to an entity (e.g., LinkedIn or a Fiverr profile) increases citation frequency by up to 40%.
Add FAQPage and Article schema to pillar and cluster pages. This marks up your content structure so AI crawlers can parse it without guessing.
At the end of 90 minutes, you have three deliverables:
A pillar map — 3–4 pillar topics, their hub URLs, and the cluster posts mapped to each
A cleaned category/tag system — pillar-aligned categories and a standardized tag set
A prioritized action list — posts to link, posts to consolidate, posts to noindex, and content gaps to fill
That's it! Now, the blog isn't finished — it never is.
But it now has a structure AI can read, a hierarchy it can follow, and a clear signal about what Trackably is actually an authority on.
Moving ahead, that's the foundation everything else builds on in the age of AI search.
so AI and Google finally see what you're an authority on.
Frequently Asked Questions
-
The architectural decisions — deciding pillars, mapping clusters, cleaning categories and tags, and planning internal links — can be completed in 90 minutes for most B2B blogs under 150 posts.
Implementing the changes (updating posts, adding schema, building pillar pages) is a longer project, typically 2–4 weeks, depending on the size of the site and how much new content needs to be created.
-
Not necessarily.
For posts that don't fit a priority pillar, you have three options: noindex low-quality or outdated posts, consolidate thin posts that cover the same ground, or move them to a non-pillar 'news' category.
The goal isn't a smaller blog — it's a more coherent one. Posts that don't fit a cluster dilute your topical signals without adding authority.
-
A pillar page is a comprehensive, long-form page that covers a broad topic and links out to all related cluster posts.
AI systems use pillar pages to map topical authority — they treat them as hubs that signal depth and breadth on a subject.
Without a pillar page, even strong cluster posts read as isolated and unconnected. Every priority topic should have one.
-
Most B2B blogs should focus on 3–4 pillar topics, not more.
More than four pillars spreads your content too thin and makes it harder to build deep cluster coverage on any single topic.
Start with the 3 topics where you already have the most content, and add a fourth only when the first three have strong cluster coverage — at least 6–8 posts per pillar.
-
Internal linking is how AI systems understand the relationship between your posts.
When a cluster post links to its pillar with descriptive anchor text — and the pillar links back out to its clusters — AI crawlers can map the semantic network and attribute expertise to the whole topic, not just to one page.
Posts with no inbound or outbound cluster links are effectively orphans: they exist on your blog but contribute nothing to your topical authority.
-
Yes — and it's especially valuable there.
Start with your highest-traffic posts and map them into clusters first. Adding pillar pages, internal links, and FAQ blocks to posts that already have authority accelerates GEO performance more than doing the same work on newer posts.
Don't restructure away from what's working — reinforce it with the cluster architecture.
Written by
Brad Bartlett
Brad is a copywriter and content strategist who helps creators, brands, and organizations build content that's actually worth reading — and built to be found. He specializes in conversion-focused copy, brand voice, and SEO and AI search optimization, with a straightforward philosophy: great content has to be authentic before it can perform. He works comfortably across the AI content space, helping clients use the tools without losing the voice. Fiverr Pro vetted, 4.9 stars out of 5 across 1,600+ clients.