Topic clusters in 2026 require passage-level structure and cross-linking between cluster pages to earn citations from AI search systems, not just consolidate ranking signals for a single pillar page. Ahrefs data shows only 38% of AI Overview citations now come from pages ranking in the top 10 organic results, down from 76%, which means clusters built solely around PageRank logic leave most AI citation opportunities untouched.
The structural gap shows up in audits: pillar pages that rank well organically but contain no self-contained, retrievable passages get skipped by AI systems pulling answers from pages with tighter passage-level structure. The teams that restructure existing clusters for passage-level retrieval will capture the citation traffic that traditional-only clusters forfeit.
Key Takeaways
- Passage structure is the upgrade. Topic clusters in 2026 must be optimized for passage-level AI retrieval, not just PageRank consolidation.
- AI citations diverge from rankings. Only 38% of AI Overview citations come from top-10 organic results (Ahrefs, 863K keyword study). Cluster pages outside the top 10 get cited when their passages answer specific sub-queries.
- Cross-linking between cluster pages matters. Hub-and-spoke alone is incomplete. Links between related cluster pages create the semantic connections that both ranking systems and AI retrieval use to evaluate topical depth.
- Self-contained passages beat comprehensive length. Five passages that each answer a specific sub-query outperform one long narrative section covering the same ground.
- Existing clusters can be retrofitted. The action is restructuring existing content at the passage level and adding cross-links, not rebuilding from scratch.
What Changed About Topic Clusters Between 2023 and 2026?
Three shifts broke the assumptions traditional topic clusters were built on: Google introduced a topic authority system, AI search systems began retrieving at the passage level instead of the page level, and AI Overview citations decoupled from organic ranking position. The old hub-and-spoke model still consolidates ranking signals. It no longer covers the structural requirements for AI citation.
If your old pillar pages are not performing the way they used to, these three changes explain why.
How Does Google’s Topic Authority System Affect Clusters?
Google’s topic authority system, introduced in May 2023, evaluates how much expertise a publication’s expertise within a specific topic area. The system looks at signals including how notable the source is for that topic, whether its original reporting gets cited by other publishers, and its history of producing high-quality content on that subject.
Google’s documentation specifically references news-related queries. But the underlying logic of evaluating topical depth applies broadly to how ranking systems assess content clusters. A cluster that covers a topic across multiple pages with original, specific content sends exactly the kind of signal this system rewards.
Why Does Passage-Level Retrieval Change the Cluster Model?
AI search systems built on retrieval-augmented generation (RAG) do not retrieve whole pages. They decompose a user’s query into sub-queries and retrieve individual passages that answer each sub-query independently. The system then synthesizes a response from the strongest passages across multiple sources.
A pillar page designed to consolidate PageRank is not the same thing as content structured for passage-level retrieval. The pillar may rank, but if its sections are not self-contained, AI systems skip it in favour of pages where each H2 opens with a direct, extractable answer. Google’s helpful content guidance reinforces this: content should demonstrate first-hand expertise at the level of the individual passage, not just the page.
How Far Have AI Citations Diverged from Organic Rankings?
The gap between organic ranking and AI citation is now measurable. Ahrefs analyzed 863,000 keyword SERPs and 4 million AI Overview URLs and found that only 38% of AI Overview citations come from pages ranking in the top 10 organic results. That figure was 76% in their earlier analysis. The remaining citations are split almost evenly: 31.2% from positions 11 through 100 and 31.0% from beyond position 100.
BrightEdge’s ongoing tracking adds an important layer. Citation overlap with organic rankings grew from 32.3% to 54.5% overall, but with enormous industry variation. Healthcare, insurance, and education show 68 to 75% overlap because trust-sensitive verticals still favour ranked content. E-commerce saw nearly flat overlap change. In YMYL and expertise-driven verticals, topic authority and ranking position still reinforce AI citation. In other verticals, passage relevance dominates.
This data reframes what the difference between traditional search engines and answer engines means in practice. Ranking is one signal. Passage structure is another. Clusters need both.
What Is the Difference Between Traditional Clusters and Passage-Optimized Clusters?
Traditional topic clusters consolidate link equity through a hub-and-spoke architecture designed to rank a pillar page for a head term. Passage-optimized clusters add a second layer: every page in the cluster contains self-contained passages structured for independent retrieval by AI systems, and the cluster pages link to one another rather than just to the pillar.
The structural differences are concrete:
| Dimension | Traditional Cluster | Passage-Optimized Cluster |
| Primary unit | Page | Passage (self-contained section under H2/H3) |
| Link architecture | Hub-and-spoke (pillar to clusters) | Hub-and-spoke + cross-linking between cluster pages |
| Content depth | Comprehensive pillar, supporting clusters | Every page has self-contained citable passages |
| Query targeting | Head term on pillar, long-tail on clusters | Sub-query coverage distributed across all pages |
| Success metric | Pillar ranking + traffic | Citation rate + sub-query coverage + traffic |
A cluster optimized for pillar ranking provides AI systems with a single retrieval entry point. A passage-optimized cluster with cross-links gives them one entry point per cluster page. That is the difference between being cited once and being cited across a topic.
Why Does Cross-Linking Between Cluster Pages Matter?
Most cluster advice focuses on pillar-to-cluster links. That is only half the architecture. Links between related cluster pages create the semantic connections that both ranking systems and AI retrieval use to evaluate topical authority. Google’s internal linking documentation confirms this: links help search engines discover pages and understand relationships between content.
Cross-links tell the system that these pages belong together and reinforce each other’s claims. Without them, each cluster page exists in partial isolation. The pillar connects them, but the individual pages do not support each other directly. That gap costs citation opportunities.
How Do You Structure Cluster Pages for AI Citation?
Each cluster page must contain at least two to three self-contained passages that answer specific sub-queries without requiring context from any other section or page. The passage opens with a direct answer under a heading that matches how a real person would phrase the question, followed by supporting evidence or a specific example.
This is why a 3,000-word guide that does not rank often fails: it covers the topic, but no individual passage stands on its own as a citable answer. Length without a passage structure is not depth. It is dilution.
What Makes a Passage Citable by AI Systems?
A citable passage meets four requirements. It is self-contained, meaning no pronouns that require surrounding context to resolve. It opens with a direct answer to a specific question. It contains specific data, examples, or expert insight not available in competing content. And it sits under a clear H2 or H3 heading that matches how someone would actually search for that information.
These are the same structural principles behind writing for AI answer engines. The format serves both traditional passage ranking and AI retrieval.
How Many Cluster Pages Should a Topic Cluster Have?
HubSpot reports that sites implementing topic clusters saw an average 43% increase in organic traffic and recommends 8 to 12 cluster pages per pillar topic. That range remains a reasonable baseline.
The 2026 adjustment: size the cluster to cover all retrievable sub-queries, not just keyword variations, and use query fan-out logic* to identify the sub-questions AI systems will decompose from the head term. If there are 15 distinct sub-queries and you have 6 cluster pages, you have 9 citation gaps. The cluster size should follow the query map, not an arbitrary number.
*Query fan-out logic: the process by which AI search systems decompose a single user query into multiple smaller sub-queries to retrieve the most relevant information.
How Do You Plan Sub-Query Coverage Across a Cluster?
Start by mapping every sub-query an AI system could generate from the primary topic. Tools like People Also Ask, Ahrefs questions reports, and manual query testing in ChatGPT and Perplexity reveal the sub-query landscape. Assign each sub-query to a specific cluster page and a specific passage within that page.
Every passage becomes an independent retrieval target. Every gap in sub-query coverage is a gap in citation potential. The goal is full sub-query coverage across the cluster, not comprehensive coverage on any single page.
How Do You Audit an Existing Cluster for AI Readiness?
Audit existing clusters by testing each page for passage-level retrievability: does every H2 section open with a self-contained answer?, and does the cluster as a whole cover the sub-queries an AI system would generate from the head term? The gap between your current structure and a passage-optimized structure tells you what to restructure first.
What Does an AI-Readiness Audit Check?
The audit runs in five steps:
- List every page in the cluster and its primary H2 sections.
- Test each H2 opener: can the first two to three sentences stand alone as a complete answer without context from other sections?
- Map sub-query coverage: which sub-queries does the cluster currently answer, and which are missing?
- Check cross-linking: are cluster pages linked to each other, or only to the pillar?
- Assess freshness: Search Engine Journal reports that 70% of pages cited in AI Overviews changed over a two-to-three-month period, which means stale passage content loses citation eligibility quickly.
The audit produces a prioritized action list: passages to rewrite, cross-links to add, sub-query gaps to fill with new cluster pages. This is how the evolution of the SEO strategy plays out at the content-architecture level.
Should You Rebuild Clusters or Retrofit Them?
The question teams keep asking is whether it is worth updating old content or just writing new stuff. The answer: retrofit first. Most teams already have clusters. The action is restructuring existing content at the passage level, not tearing down what already ranks.
Restructure the pillar page so each H2 contains a self-contained passage. Add cross-links between cluster pages. Fill sub-query gaps with targeted new cluster pages. This preserves existing ranking signals while adding the passage-level structure that AI retrieval requires.
Do Topic Clusters Still Work for SEO in 2026?
Topic clusters still work. They consolidate ranking signals, build topical authority, and drive organic traffic. The 2026 shift is not away from clusters but toward clusters that serve both traditional ranking and AI citation by adding passage-level structure and cross-linking.
The data confirms it. HubSpot’s findings on organic traffic growth from clusters have not been contradicted. BrightEdge data shows that in YMYL verticals, 68 to 75% citation overlap with rankings means strong clusters benefit from both signals simultaneously.
The teams that ask “Do we need to restructure everything for AI search?” are asking the wrong question. You do not need to restructure everything. You need to add passage-level structure to your existing clusters and connect cluster pages to one another. That is an upgrade, not a replacement.
Teams that treat this as a binary choice between traditional SEO and answer engine optimization will underperform those that build for both. The cluster model is not broken. It is incomplete. The fix is structural, not strategic. And it starts with the next cluster you audit.
Frequently Asked Questions
What are topic clusters in SEO?
A topic cluster is a content architecture where a pillar page covers a broad topic and links to cluster pages that cover related subtopics in depth. The cluster pages link back to the pillar and to each other. The structure helps search engines understand topical authority and helps AI systems retrieve passages across the full topic.
How do topic clusters help with AI search visibility?
Topic clusters give AI systems multiple retrieval entry points across a topic instead of relying on a single page. When each cluster page contains self-contained passages that answer specific sub-queries, AI search systems can cite the most relevant passage from any page in the cluster.
What is the difference between a pillar page and a cluster page?
A pillar page covers a broad topic comprehensively and serves as the central hub of a topic cluster. Cluster pages cover specific subtopics in depth and link back to the pillar. In a passage-optimized cluster, both pillar and cluster pages must contain self-contained passages structured for independent retrieval.
How many cluster pages should a topic cluster have?
HubSpot recommends 8 to 12 cluster pages per pillar topic. The right number depends on how many distinct sub-queries exist for your topic. Size the cluster to cover every sub-query an AI system could generate from the head term, not just keyword variations.
Do topic clusters still work for SEO in 2026?
Topic clusters remain effective for consolidating ranking signals and building topical authority. The 2026 adjustment is adding passage-level structure and cross-linking between cluster pages, so the same architecture also earns AI citations. The model is an upgrade, not a replacement.
Structuring your content for both traditional ranking and AI citation is not a future problem. It is happening now, and the teams that adapt their cluster architecture will own the retrieval layer that single-page strategies cannot reach. Blacksmith SEO works with marketing teams and content leads to build content systems that perform across both channels. If you are ready to upgrade your cluster strategy, start a conversation.
