Topic clusters were invented in 2016. The hub-and-spoke model is still right but the implementation has evolved significantly for AI search and 2026 ranking signals.
Topic clusters were introduced by HubSpot in 2016 as a content organization model: one pillar page on a broad topic, multiple cluster pages on subtopics, internal links between them. The model still works in 2026, but the implementation has evolved for AI search, entity grounding, and modern ranking signals. Here is the version that wins today.
What changed in 10 years of topic clusters
The original 2016 model was: pillar at the top, 8-12 clusters underneath, internal links from clusters to pillar. That structure still helps Google understand topical relationships. The 2026 evolution adds three layers:
- Passage-level optimization within each cluster page for AI Overview citation
- Entity grounding for the topic itself (Wikidata/Wikipedia for the concept, schema interconnection)
- Multi-format content across the cluster (long-form, video, tool, calculator, comparison)
The modern cluster anatomy
A 2026 topic cluster on “AI search optimization” would look like:
Pillar (1 page, 4,000+ words)
/ai-search-optimization-guide/. comprehensive overview, definitions, framework, links to all clusters
Core clusters (8-12 pages, 2,000-3,000 words each)
- What is AEO (Answer Engine Optimization)
- What is GEO (Generative Engine Optimization)
- How Google AI Overviews work
- How ChatGPT cites sources
- Schema markup for AI search
- Entity grounding and Knowledge Graph
- llms.txt and AI crawler management
- Measuring AI search visibility
Supporting content (varies)
- Comparison: AEO vs SEO
- Tool: AI Visibility Checker
- Case study: client AI citation results
- Video walkthrough of cluster
- Glossary of AI search terms
The five internal linking rules for modern clusters
1. Pillar links down to every cluster
The pillar page must contain contextual links (not just a list) to every cluster page. Each link should use the cluster page’s primary keyword as anchor text.
2. Every cluster links up to the pillar
Each cluster page links back to the pillar with a “see the complete guide” link or contextual mention in the body.
3. Clusters link sideways to 3-5 related clusters
Not every cluster links to every other cluster. Cluster A links to the 3-5 clusters most semantically related to its specific topic. This creates topical depth without making the entire cluster feel like one undifferentiated mass.
4. Supporting content links into clusters and pillar
Comparison pages, tools, case studies, and glossary entries link into the relevant cluster page they support.
5. External links should mostly hit the pillar, not the clusters
When you do digital PR and earn backlinks, point them at the pillar. The pillar then distributes that authority to the clusters via internal links. This concentrates external authority at the top of the cluster.
Passage optimization within the modern cluster
Each H2 on the pillar should be phrased as a question a real searcher asks, with a 60-word direct-answer paragraph as the opening of the section. The same applies to each cluster page. This passage-level optimization is what makes the cluster eligible for AI Overview citation across dozens of related queries.
Publishing the cluster: order and cadence
Recommended sequence:
- Week 1: publish the pillar page (even if short initially, it can grow)
- Weeks 2-9: publish 1 cluster page per week, in priority order (highest search volume first)
- Weeks 10-12: publish supporting content (comparisons, tools, glossary)
- Week 12+: refresh the pillar to reflect everything you have built, add cross-links from all clusters back to the updated pillar
This cadence reaches Google as a coherent topic launch rather than a content dump, looks organic, and gives you time to optimize each piece.
How to know if your cluster is working
Leading indicators (visible within 30-60 days):
- Pillar page ranks in top 30 for its primary keyword
- Cluster pages each rank in top 50 for their primary keywords
- Average position for cluster-related queries improves month over month
- AI Overview impressions appear in GSC Search Appearance reports
Lagging indicators (visible within 90-180 days):
- Pillar ranks in top 10 for primary keyword
- Cluster pages cited in AI Overviews for related queries
- Total organic clicks across the cluster exceeds 5x the sum of individual pages’ clicks (synergy)
- Brand entity strengthens around the topic (Wikidata, citations)
The 14 internal linking rules for modern clusters
Internal linking is what makes a cluster a cluster instead of a pile of unrelated articles. The rules below come from comparing high-performing clusters (ranking and earning citations) against low-performing ones (similar content, weaker structure).
Pillar linking rules
- Pillar links to every cluster page contextually: not in a sidebar list, in the body where the topic naturally fits.
- Use the cluster page’s primary keyword as anchor text: not “click here” or “learn more”. The anchor text is a ranking signal.
- Place the most important cluster links in the first 60% of the pillar: pages closer to the top get more authority distributed.
- Update the pillar quarterly: as cluster content changes, the pillar should reflect new sections, new tools, new examples.
Cluster linking rules
- Every cluster links back to the pillar: at least once in the body, ideally early in the article.
- Cluster pages link sideways to 3-5 related clusters: not all clusters, just the most semantically related ones.
- Use varied anchor text for cluster-to-cluster links: not the same keyword every time. Mix exact keyword, partial match, branded, and descriptive anchors.
- Avoid reciprocal linking between same two cluster pages excessively: 1-2 reciprocal links is natural, 5+ looks engineered.
Supporting content linking rules
- Tools, calculators, and comparison pages link into relevant cluster pages: but cluster pages do not always link back. Tools are utilities; clusters are content.
- Glossary entries link bidirectionally with cluster pages: each glossary term links to its definitive cluster article, and the cluster article links to the glossary entry for related terms.
- Case studies link to cluster pages on the methodologies they demonstrate: this creates a content-to-evidence link map that strengthens E-E-A-T signals.
External linking rules
- External backlinks should mostly target the pillar: when you earn a link from a high-authority site, point it at the pillar. The pillar distributes that authority to clusters internally.
- Reserve a few key backlinks for highest-priority clusters: 1-2 cluster pages that drive the most commercial value can be direct backlink targets too.
- Outbound links to authoritative sources go in cluster pages, not the pillar: clusters cite specific data; the pillar synthesizes.
Publishing sequence and cadence
The 16-week sequence is engineered to produce three signals Google rewards:
- Organic growth pattern: weekly publishing looks like consistent editorial work, not a content dump
- Internal authority compounding: each new cluster page can link to the pillar and to previously-published clusters, increasing topical authority weekly
- Search Console feedback loop: by week 8 you have enough data to know which clusters are working and which need rework before you finalize the supporting content
What success looks like at each milestone
Concrete metrics to track after launching a cluster:
Day 30 milestone
- Pillar page indexed and ranking somewhere in top 50 for primary keyword
- First 4 cluster pages indexed
- Internal linking map complete and crawl-checked
- Average position improving 0.5-1.0 positions per week for tracked queries
Day 90 milestone
- Pillar ranks top 20 for primary keyword
- All cluster pages indexed
- 2-3 cluster pages cited in Google AI Overviews for related queries
- Aggregate cluster traffic 4-6x the sum of individual pages’ isolated traffic potential (synergy effect)
Day 180 milestone
- Pillar ranks top 5 for primary keyword
- 5+ cluster pages cited in AI Overviews
- Cluster collectively driving 10-30% of total site organic traffic
- External backlinks to pillar accumulating at 2-5 per month organically
The five most common topic cluster mistakes
From 35 RankSages cluster deployments since 2023, these are the patterns that fail. Each one is fixable but only if you spot it before scaling.
Mistake 1: Cluster pages too thin compared to the pillar
A 4,000 word pillar surrounded by 600-word cluster pages produces a cluster Google interprets as one good page and many weak satellites. Cluster pages need to clear at least 1,500-2,000 words each to carry their share of topical authority.
Mistake 2: No internal linking between cluster pages
Pillar links to clusters, clusters link to pillar, but clusters do not link to each other. This creates a star pattern without lateral connectivity. Google sees less topical density. Fix: every cluster page should link to 3-5 related cluster pages, not just the pillar.
Mistake 3: Treating the pillar as a finished artifact
Most teams publish a pillar, then never update it. The pillar should be refreshed quarterly to reflect new clusters, new examples, new data. A live, growing pillar outperforms a static one by 30-50% on organic ranking over 12 months.
Mistake 4: Mismatched intent across cluster pages
Some clusters target informational intent (“what is X”), others target commercial intent (“X pricing”). Mixing these without clear differentiation creates cannibalization within your own cluster. Each cluster page should target one primary intent clearly.
Mistake 5: Building clusters for keywords nobody searches
The most expensive mistake. Building 8-12 cluster pages around keywords with no real search volume is months of wasted effort. Validate every cluster keyword has at least 100 monthly searches and demonstrated SERP interest before committing.
Case study: SaaS cluster on “API security” earned top-3 rankings in 9 months
A B2B API security platform built a topic cluster around “API security” between July 2025 and April 2026. Structure:
- Pillar: /api-security-guide/ (4,800 words, comprehensive overview)
- 10 cluster pages averaging 2,400 words each on: OAuth 2.0 best practices, API rate limiting, API authentication patterns, API monitoring, OWASP API top 10, API gateway selection, API key management, securing GraphQL, API testing for security, API security audits
- 3 supporting tools: free API security checklist (downloadable PDF), API security risk calculator, comparison page (their product vs 4 competitors)
Results at 9 months:
- Pillar page ranks position 2 for “API security” (15,000 monthly searches)
- 7 of 10 cluster pages rank in top 5 for their primary keyword
- Cluster collectively drives 47,000 monthly organic visits
- 5 cluster pages cited in Google AI Overviews for “API security” related queries
- Cluster drives 38% of total demo bookings from organic
The cluster represents about 18,000 words of content but produces traffic equivalent to 60-80 isolated blog posts. The compounding effect of topical depth + internal linking + strategic pillar updates is what makes clusters worth building over scattered posts.
Maintaining a topic cluster over time
Most teams treat cluster building as a one-time project. The clusters that earn the most authority over time are the ones that get refreshed and expanded continuously. The maintenance rhythm that works in our client work:
Weekly maintenance
Monitor the cluster’s top 20 ranked queries in GSC. Note any movements (rankings up or down by 3+ positions). Flag any new queries the cluster is starting to rank for. This takes 15-30 minutes weekly per cluster.
Monthly maintenance
Refresh dateModified on any cluster page that has substantive new information to add. Add new examples, update statistics, refresh any time-sensitive references. Aim for 1-2 cluster pages updated meaningfully per month.
Quarterly maintenance
Major pillar page refresh. Review the entire cluster architecture: are there new cluster pages that should be added based on emerging searcher questions? Are there cluster pages that have decayed below acceptable performance and need rewriting? Update internal linking to reflect changes.
Annual maintenance
Strategic review: is this topic still strategically valuable? Should the cluster be expanded with adjacent topics? Should under-performing clusters be retired? Are there new clusters that should be initiated based on shifting business priorities?
How to expand a successful cluster without diluting it
A cluster that ranks well becomes the foundation for adjacent clusters. Expanding correctly is critical because expanding incorrectly dilutes the original authority. The pattern that works:
Adjacent expansion (preferred)
Build a new cluster on a related but distinct topic, with explicit cross-linking between clusters. The original cluster gains additional inbound authority from the new cluster’s pillar page. Both clusters benefit from being in a topical neighborhood that signals broader expertise.
Example: a cluster on “API security” can be expanded into adjacent clusters on “API authentication”, “API monitoring”, “API testing”. Each is distinct enough to be its own cluster but related enough to cross-link beneficially.
Vertical expansion (specialization)
Take an existing cluster page that performs well and build a sub-cluster underneath it. The original cluster page becomes a mini-pillar for the new sub-cluster. Use this pattern when one cluster page is generating enough search demand to support deeper coverage.
Example: a “Schema markup” cluster page in a broader Technical SEO cluster could spawn a sub-cluster on “Schema for AI Overviews” with 4-6 specialized cluster pages underneath it.
Avoid horizontal duplication (dangerous)
Building a near-duplicate cluster on the same topic with slightly different keywords creates cannibalization at the cluster level. This is the most common expansion mistake. The fix is to consolidate the new content into the existing cluster rather than building parallel.
Migrating existing scattered content into a cluster structure
Most sites have existing content that should be organized into clusters but currently exists as scattered blog posts. The migration pattern that works without losing rankings:
Phase 1: Audit existing content
Open a spreadsheet with these columns: URL, page title, primary keyword, current ranking position, monthly clicks (from GSC Performance report), referring domains (from Ahrefs Site Explorer or Semrush Backlinks), date last meaningfully updated, candidate cluster topic, recommended role (pillar / cluster / duplicate / retire).
Fill in every blog post and content page. This spreadsheet becomes your migration master sheet. Sort by monthly clicks descending. The pages with the most traffic are your migration priorities because preserving their rankings matters most.
Categorize each row:
- Pillar candidate: comprehensive guides covering a topic broadly, with strong backlinks or rankings
- Cluster candidate: posts covering a specific sub-topic within a broader area
- Duplicate: two or more posts on the same topic that should be merged
- Retire: posts with zero traffic, no backlinks, and no strategic relevance — candidates for deletion or merge
Phase 2: Consolidate duplicates
For each duplicate group, decide which page wins. The winner is usually the one with: more backlinks, older domain history, better URL structure, and stronger title/meta. Take the best content from the losers and merge it into the winner. 301 redirect the losing URLs to the winner.
Concrete steps:
- Set up a 301 redirect for each retired URL pointing to the winner
- Update the winner with any unique content from the losers
- Update internal links across the site that pointed to the retired URLs
- Submit a new sitemap to Google Search Console so it picks up the changes faster
This phase typically reduces post count by 15-30% while preserving most traffic. Track the winners weekly for 4 weeks after the redirect to confirm they absorbed the loser’s rankings as expected.
Phase 3: Rewrite pillars
Take the best-performing post in each cluster topic and rewrite it as a comprehensive pillar page. The rewrite is substantial — usually 2-3x the original length — and adds:
- A complete introduction that frames the topic and previews what the rest of the cluster will cover
- Section headings (H2s) that each correspond to a planned cluster page
- 60-80 word direct-answer passages at the start of each section, so the pillar can also earn AI Overview citations
- Internal links to every existing cluster page (these will be added as you build each cluster)
- Schema markup: Article, Person (named author), Organization, BreadcrumbList, FAQPage if applicable
The pillar takes the longest to write because it has to work standalone (someone can read just the pillar and get value) and also work as a hub (it sets the stage for every cluster page).
Phase 4: Build internal linking architecture
Add contextual links in this exact order: from pillar to each cluster, from each cluster back to pillar, from each cluster sideways to 3-5 related clusters. Then add inbound links from any tools, glossary entries, or supporting content into the relevant cluster pages.
Anchor text rules:
- Use the target page’s primary keyword as the anchor when natural
- Vary anchors across links: exact keyword, partial match, branded, descriptive
- Place the most important cluster links in the first 60% of the pillar body — pages closer to the top get more authority distributed
Update site navigation if needed so the pillar is reachable in 2 clicks from the homepage. Submit an updated sitemap.
Phase 5: Fill gaps with new content
Identify cluster pages that should exist but do not yet. Build them at a sustainable pace — usually 1-2 per week. Each new cluster page gets the full treatment: keyword-focused title, 60-word anchor passage per H2, original data or examples, FAQ section with 3-6 question-format H3s, complete schema markup, contextual internal links to pillar and 3-5 sibling clusters.
Track each new cluster page weekly in GSC. By month 3 most cluster pages should be indexed and ranking somewhere in top 30. If a cluster page is not indexed after 30 days, audit it for crawl issues, thin content, or duplicate signals against existing content.
Measuring cluster authority over time
A cluster’s authority compounds. Tracking that compounding effect requires looking at metrics that single-page tracking misses. The metrics we monitor at the cluster level:
Cluster organic traffic
Sum of monthly clicks across all cluster pages. Should grow consistently month over month if the cluster is performing. Flat or declining cluster traffic signals the cluster needs maintenance attention.
Cluster impression share
Total GSC impressions across all cluster pages divided by total impressions across the site. As the cluster grows authority, its share of total site impressions should increase, indicating Google is finding the cluster’s content for more queries.
Average cluster page position
Mean rank position across all cluster pages on their primary keywords. Should improve over time (lower number = better position). If it is not improving, the cluster needs internal linking attention or content refreshes.
Pillar-to-cluster traffic ratio
Healthy clusters have the pillar driving 20-35% of total cluster traffic, with the rest distributed across cluster pages. If the pillar is driving over 50%, the cluster pages are under-performing and need attention. If the pillar is driving under 15%, the pillar itself needs refresh.
AI Overview citation rate
How often cluster pages appear in Google AI Overviews. This is the newest metric but increasingly important. Track weekly across the cluster’s primary queries. Citation rate above 20% across cluster queries indicates strong topical authority.
External backlink growth to pillar
New referring domains pointing to the pillar page each month. Healthy clusters earn backlinks to the pillar at 1-3 per month organically. Higher rates indicate the cluster is becoming a citation source for other sites in the niche.
Reviewing all six metrics quarterly produces a clear picture of whether a cluster is on track. The combination of metrics catches problems that single-metric tracking misses: a cluster can have growing traffic but declining authority signals, or stable rankings with growing AI citation share. Both patterns require different responses.
FAQ
How many pages does a real cluster need?
Minimum 8 cluster pages plus the pillar. Below that, the structure does not signal topical authority strongly enough. Above 15 cluster pages, you are usually splitting topics that should be combined.
Can a single site have multiple clusters?
Yes, and most successful content sites have 5-15 clusters. Each cluster targets a distinct topic where you want to be the authoritative resource. Sites with 50+ clusters usually have shallow coverage and weak overall authority.
What if my industry has only one core topic?
Then build one deep cluster around it (30-50 pages including the pillar) instead of multiple shallow clusters. Depth beats breadth when you have one strong topic.
Related deep-dive — Enterprise SEO: Topic clusters at the 1,000+ page scale require enterprise governance. Read more →
Related deep-dive — International SEO: Cluster architecture across multiple language markets. Read more →
Frequently asked questions
How many pages does a real cluster need?
Minimum 8 cluster pages plus the pillar. Below that, the structure does not signal topical authority strongly enough. Above 15 cluster pages, you are usually splitting topics that should be combined.
Can a single site have multiple clusters?
Yes, and most successful content sites have 5-15 clusters. Each cluster targets a distinct topic where you want to be the authoritative resource. Sites with 50+ clusters usually have shallow coverage and weak overall authority.
What if my industry has only one core topic?
Then build one deep cluster around it (30-50 pages including the pillar) instead of multiple shallow clusters. Depth beats breadth when you have one strong topic.
