First published: 15 June 2026 · Last updated: 15 June 2026
(broad topic, 3-5K words)
subtopic deep-dive
subtopic deep-dive
subtopic deep-dive
subtopic deep-dive
subtopic deep-dive
subtopic deep-dive
subtopic deep-dive
subtopic deep-dive
The conversation about topical authority has become noisier in 2026 than it was in 2022. Vendor blog posts now use "topical authority" as a synonym for "publishing many articles on a topic", which is wrong. Genuine topical authority is structural: a coherent site architecture that signals to Google and AI engines that your site is a credible source on a specific topic space, demonstrated through entity coverage, internal linking, and depth-per-cluster. Sites that publish 50 articles on a topic with no architecture rarely build topical authority. Sites that publish 12 articles on a topic with mature hub-and-spoke architecture build authority that compounds for years.
This article is the practitioner's view of the modern architecture. We work through the hub-and-spoke model, the internal linking patterns Google rewards in 2026, semantic and entity clustering, and the 6-step build process for a new cluster. The architecture connects directly to our broader E-E-A-T 2026 framework (topical authority is the structural component of expertise) and our GEO optimization tactics playbook (topical authority is what AI engines use to assess credibility before citing you). For the foundational SEO content principles we still lean on our SEO-friendly content guide; this article layers architecture on top.
Why Architecture Beats Volume
The 2026 SEO data is consistent on this. Sites that achieve top rankings on competitive topics in 2026 are not the sites that publish the most content. They are the sites with the most coherent architecture. Three reasons.
Reason 1: Google evaluates topic coverage at the cluster level, not the article level. Google's quality and topic-authority signals look at how completely a site covers an entity space. A site with 12 deeply-interconnected articles on "B2B lead generation" demonstrates more topical authority than a site with 50 disconnected articles on the same topic. The cluster's coherence signals expertise; the volume alone signals content production.
Reason 2: Internal linking distributes authority and signals topic priority. Internal links from your strongest pages to cluster content tells Google "this content is important to my site". A hub-and-spoke architecture concentrates internal links efficiently. A flat architecture with no clustering wastes the link signal across hundreds of pages.
Reason 3: AI engines triangulate authority across the cluster. When Google AI Overviews or Perplexity selects a citation for a query within a topic, they assess the source page's surrounding context (other pages on the topic, depth of cluster, entity coverage). A cited page on a thin site rarely gets cited again; a cited page on a mature cluster gets cited repeatedly across related queries.
The strategic implication. If you have a finite content budget (which you always do), deploy it to build mature clusters on a small number of topics rather than thin coverage of many topics. Three mature clusters of 12 articles each (36 total) outperform one shallow library of 100 disconnected articles for both organic ranking and AI citation.
The Hub-and-Spoke Model in Detail
The hub-and-spoke (or pillar-cluster) model is the dominant 2026 architecture pattern. The components.
The Pillar (Hub)
The pillar is the broad-topic page. It covers the topic comprehensively at depth (typically 3,000 to 5,000 words), introduces the major subtopics, and links to each cluster page that covers a subtopic in depth. The pillar targets a high-volume, broad-intent keyword (e.g., "B2B Lead Generation in Singapore").
The pillar's structural responsibilities:
- Comprehensive but not exhaustive coverage of the broad topic
- Outbound contextual links to every cluster page in the cluster
- Position in the site as the canonical entry point for the topic
- Long-form structure with chunked subsections (per our long-form for AEO guide)
- Targets the broad-intent primary KW for the topic space
The Cluster (Spokes)
Each cluster page covers a specific subtopic in depth (typically 1,500 to 3,000 words). Each targets a more specific keyword within the topic space (e.g., "B2B SDR cold email Singapore", "B2B LinkedIn lead gen tactics", "B2B inbound vs outbound for SG SaaS").
The cluster page's structural responsibilities:
- Deep coverage of the specific subtopic, not just a slice of the pillar
- Inbound link from pillar with descriptive anchor text
- Outbound link back to pillar with branded or topic-broad anchor
- 2 to 4 lateral links to other cluster pages in the same cluster
- Targets a specific, often longer-tail keyword
The Recommended Cluster Size
The 2026 industry consensus is 8 to 15 cluster pages per pillar. Below 5, the cluster is too thin to demonstrate topical authority. Above 20 to 25, the cluster is large enough that Google may treat it as multiple distinct clusters (and the pillar starts losing its anchoring role). The 8 to 15 range is the sweet spot for most topics.
Exceptions. Very broad topics (e.g., "ecommerce SEO" covering 50+ subtopics) genuinely need to be split into multiple clusters with sub-pillars. Very narrow topics (e.g., "Schema for SG legal firms") may only support 4 to 6 cluster pages and that is fine; do not pad to hit 8 with thin content.
The Internal Linking Patterns Google Rewards in 2026
Internal linking is the structural mechanism that converts a content library into a topic cluster. The 2026 patterns.
Pillar → Spokes (top-down)
Pillar contextually links to every cluster page using descriptive anchor text. Each spoke gets at least one link from the pillar. The pillar serves as the cluster's canonical map.
Spokes → Pillar (bottom-up)
Every spoke links back to the pillar. Use the pillar's primary KW or a close variant as anchor text. This signals to Google that the pillar is the cluster's hub.
Spoke ↔ Spoke (lateral)
Each spoke links to 2 to 4 other spokes in the same cluster. Lateral links signal topic-coherence and distribute authority within the cluster.
Cross-cluster links (selective)
Where topics genuinely overlap, link between clusters. Avoid forcing cross-links for SEO; they should serve reader value first.
The anchor text discipline. Internal anchor text should be descriptive and natural, mixing exact-match (for the spoke's primary KW), partial-match (variants), and branded (the spoke's title) anchors. The mix:
- 20 percent exact-match (e.g., "B2B SDR cold email")
- 30 percent partial-match (e.g., "cold email tactics for B2B SDRs")
- 50 percent broad / descriptive (e.g., "our deep-dive on B2B cold email")
This mix avoids over-optimisation while still signalling topic relevance. Pure exact-match repetition is a 2026 over-optimisation signal that Google is increasingly attentive to.
The linking density. We target 5 to 7 internal links per article for cluster pages, distributed roughly:
- 1 link to the pillar (always)
- 2 to 3 links to other cluster pages (lateral)
- 1 to 2 links to relevant money pages (conversion utility)
- 1 to 2 links to cross-cluster or broader topical content (where genuinely relevant)
This density is supported by content depth; thinner pages cannot sustain 7 contextual links without feeling stuffed.
Semantic and Entity Clustering: The 2026 Shift
The 2022 model of topic clusters was keyword-based: identify the keyword space, build a cluster of articles that target each keyword. The 2026 model is entity-based: identify the entity space (the people, products, organisations, processes, and concepts that define the topic), build a cluster that demonstrates authoritative coverage of those entities.
The shift is driven by Google's increasingly entity-based topic understanding. Google's Knowledge Graph connects entities and uses them to evaluate whether your content genuinely covers a topic or just mentions relevant keywords. A B2B lead-gen cluster that covers SDR (an entity), Outreach.io (entity), Apollo (entity), HubSpot (entity), Calendly (entity), and the named SG B2B SaaS companies (entities) demonstrates topic mastery in a way that a keyword-only cluster does not.
How to identify the entity space
The practical workflow:
- List the people associated with the topic. Authors, thought leaders, practitioners, executives in the space. For B2B lead gen: Aaron Ross, Jason Lemkin, Pete Kazanjy, etc.
- List the organisations and products. Tools, vendors, agencies, key brands. For B2B lead gen: Outreach, Apollo, Salesloft, ZoomInfo, HubSpot, Salesforce.
- List the processes and methodologies. Named frameworks, sales motions, methodologies. For B2B lead gen: Predictable Revenue, BANT, MEDDIC, ABM, PLG.
- List the concepts and metrics. SQL, MQL, CAC, LTV, ICP, conversion rate.
- For SG-specific clusters, add SG entities. Local agencies, SG-based SaaS companies, SG market events, SG-specific regulations.
The cluster then needs to cover these entities meaningfully across its articles. Not every article needs to cover every entity, but the cluster as a whole should demonstrate that you can discuss the entity space credibly.
How to demonstrate entity coverage
Practical patterns:
- Mention named entities by name in body content (not just in passing).
- Link to authoritative external sources for major entities (entity disambiguation).
- Use Person and Organization schema (per our GEO schema deep-dive) to make entity references machine-readable.
- Quote, reference, or analyse named experts in the space.
- Reference specific product features, pricing, or methodologies by name.
The implication for content briefs. Briefs for cluster articles should specify "must mention entities X, Y, Z" alongside the keyword targets. This forces entity coverage as a writing discipline rather than leaving it to chance.
The 6-Step Process for Building a New Cluster
The process we run when building a new topic cluster from scratch.
Topic and entity space mapping
Define the broad topic. List the entities (people, orgs, processes, concepts) that define it. Identify which entities your site can credibly cover.
Pillar and spoke planning
Define the pillar's primary KW (broad-intent). List 8 to 15 spoke topics with their primary KWs (longer-tail). Validate KW volume and intent.
Pillar build and ship
Write the pillar first as the cluster's anchor. 3,000 to 5,000 words covering the broad topic comprehensively. Forward-link to spoke topics (even if spokes not yet written).
Spoke production cadence
Ship 1 to 3 spokes per month. Each spoke deep on its subtopic, links back to pillar, lateral links to 2 to 4 other shipped spokes.
Internal link backfill
As each new spoke ships, retrofit internal links from existing spokes to the new spoke (and vice versa). The cluster grows tighter over time.
Maturity audit and refresh
At 6 to 9 months, audit cluster performance. Identify spokes that ranked, spokes that did not. Refresh underperformers, add new spokes for emergent subtopics.
The cadence. A typical cluster takes 4 to 6 months to mature from pillar-first publication. The first 3 months ship the pillar plus 6 to 8 spokes. Months 4 to 6 fill in remaining spokes and tighten internal linking. Months 7 to 9 are the maturity window where rankings begin to stabilise and AI citations begin to appear.
The realistic expectation. Pillar pages typically take 6 to 9 months to reach top-3 organic rankings on competitive primary KWs. Spoke pages typically rank within 3 to 6 months on longer-tail KWs. AI Overview citations begin to appear at 4 to 8 months as the cluster reaches authority maturity.
What Goes Wrong (Common Cluster Failures)
Patterns of failure we see consistently in cluster builds.
Failure 1: Building spokes without a pillar. Producing 10 articles on subtopics with no anchoring pillar. The articles compete with each other internally and never accumulate topical authority. Always start with the pillar.
Failure 2: Pillar that is just a directory of links. Some agencies build pillars that are 800 words of generic intro plus a list of links to spokes. This wastes the pillar's authority potential. The pillar must itself be a comprehensive standalone resource (3,000 to 5,000 words).
Failure 3: Spokes that are slices of the pillar, not deeper coverage. Each spoke should go meaningfully deeper than the pillar's coverage of that subtopic. Spokes that just elaborate on pillar paragraphs are duplicative and rarely rank.
Failure 4: Missing lateral links between spokes. Spokes that only link up to pillar (no lateral links) miss the cluster-coherence signal. Always add 2 to 4 lateral links per spoke as the cluster matures.
Failure 5: Forced cross-cluster linking. Cramming links between unrelated clusters to increase internal links per page. Google's 2026 algorithms detect forced linking and discount it. Cross-cluster links should serve genuine reader value or be omitted.
Failure 6: No entity coverage discipline. Producing keyword-targeted content without specifying which entities each article must cover. The cluster ranks for keywords but does not demonstrate entity expertise to AI engines.
Failure 7: Stopping at first publication. Treating the cluster as "done" once all spokes are published. Mature clusters require ongoing internal-link backfill, periodic refreshes, and additions for emergent subtopics. Set-and-forget clusters decay within 12 to 18 months.
Site-Wide Architecture: How Clusters Fit into the Whole Site
Individual clusters live within a broader site architecture. The 2026 best practices for site-wide structure.
URL Structure
Cluster pages should live under predictable, hierarchical URLs:
- Pillar: /blog/[topic]/ or /resources/[topic]/
- Spokes: /blog/[topic-related-slug]/
Avoid overly nested URL structures (/blog/topic/subtopic/spoke-slug/) which create long URLs and risk creating false hierarchies that Google misinterprets. Flat-blog structures (all posts under /blog/) are usually cleaner than deeply-nested ones.
Navigation
The pillar should be reachable from main navigation or from a topic hub page. Spokes do not need to appear in main navigation; they are reachable via the pillar and via internal links. Adding all spokes to main navigation dilutes the navigation's signal value.
Sitemap and Crawl Priority
In your XML sitemap, set higher priority and changefreq on pillars (priority 0.8 to 1.0) than on spokes (priority 0.5 to 0.7). This signals to Google which URLs are the cluster's authority anchors.
Breadcrumbs
Use BreadcrumbList schema on every cluster page (covered in our AEO schema types guide). Breadcrumbs reinforce the topic hierarchy for Google and AI engines, and they improve user navigation through the cluster.
A Worked Cluster: 6-Month Build for a SaaS Client
Concrete worked example. Client: B2B SaaS in HR tech, building topical authority around "HRIS for SG SMEs". Starting state: 4 disconnected blog posts, no clear topical structure.
The cluster build over 6 months:
Month 1: Pillar built and shipped. "HRIS for SG SMEs: The 2026 Guide" at 4,200 words. Comprehensive coverage of the topic with forward-links to 12 planned spoke topics.
Months 2-3: 6 spokes shipped. Each spoke covers a specific subtopic (HRIS pricing in SG, HRIS integration with payroll, HRIS for sub-50 employee SMEs, HRIS data security in SG, HRIS implementation timeline, HRIS ROI calculation). Each spoke links to pillar plus 2 lateral links to other shipped spokes.
Months 4-5: 5 more spokes shipped (HRIS vs payroll-only systems, HRIS for hybrid SG/MY teams, HRIS reporting capabilities, HRIS user experience review, HRIS migration from legacy systems). Lateral links retrofitted across all 11 shipped spokes.
Month 6: 1 final spoke shipped, cluster maturity audit, internal link backfill across all 12 spokes plus pillar. Cluster total: 13 pages (1 pillar + 12 spokes), 38,000 words combined.
Outcomes at month 9 (3 months post-cluster-completion):
- Pillar primary KW: rank position 4 (from baseline outside top 100)
- Spoke pages: 8 of 12 ranked top 10 for their primary KWs
- Organic traffic to cluster: +280% versus the pre-cluster baseline of 4 disconnected posts on similar topics
- AI Overview citations: detected on 14 of 30 sampled related queries (from baseline 0 to 1)
- Cluster total inbound backlinks: 47 (from baseline 6); the cluster's coherence and depth attracted natural backlinks
The pattern holds across our portfolio. Clusters that reach maturity at 6 months produce measurable rank lifts at 9 months and durable authority that sustains for 18 to 24 months without major refresh.
Frequently Asked Questions
How many cluster pages should a pillar have?
8 to 15 cluster pages per pillar is the 2026 sweet spot. Below 5 spokes, the cluster is too thin to demonstrate topical authority. Above 20 to 25, the cluster either needs to split into multiple sub-pillars or risks losing the pillar's anchoring role. Specific topics may warrant exceptions: very narrow topics may only support 4 to 6 spokes (and that is fine if the spokes are deep), very broad topics may require splitting into multiple clusters.
Should I build all spokes before publishing the pillar?
No. Build and ship the pillar first as the cluster's anchor, even before any spokes exist. The pillar can forward-link to planned spoke topics (using internal anchor URLs that will be valid once spokes ship). This sequence lets the pillar accumulate authority during the months you build out the cluster, instead of waiting until everything is ready.
How do I handle existing disconnected content I already have?
Two options. Option 1: identify which existing posts could fit as spokes in a new cluster, then build a pillar around them and retrofit internal links. Option 2: archive or consolidate weak existing posts into a cleaner cluster. The right choice depends on the quality of the existing posts. Strong posts can usually be retrofitted into clusters; thin posts are usually better consolidated into deeper pillar coverage.
Does cluster architecture matter for AI citations as well as Google ranking?
Yes, possibly more. AI engines use cluster depth as an authority proxy when selecting citation sources. A cited page on a mature cluster gets cited repeatedly across related queries; a cited page on a thin site rarely gets re-cited. Our portfolio data shows roughly 3.2x more AI citations on cluster-architected content versus standalone content, larger than the organic ranking lift.
How often should I refresh cluster content?
Quarterly review of pillar performance, semi-annual review of spokes. Refresh triggers include: ranking decline of more than 10 positions, AI Overview citation loss, content that has become factually outdated, new entities that should be covered. Mature clusters need ongoing maintenance; set-and-forget clusters decay within 12 to 18 months as the topic landscape evolves.
Can a cluster span multiple money pages?
Yes, often this is desirable. A cluster on "B2B SEO in Singapore" might link to money pages including SEO services, B2B lead generation, content marketing, and analytics. The pattern: each spoke links to the most contextually relevant money page (1 to 2 money page links per spoke) plus the pillar links to the primary money page (the SEO service page). This converts cluster authority into revenue across multiple service lines, not just one.