Topic clusters in 2026: what still works, what doesn't, what to do instead

Topic clusters became the default content strategy somewhere around 2019. Pillar pages plus supporting cluster pages, all interlinked, demonstrating topical authority to search engines. The strategy worked. It still mostly works. But several specific tactics that came along with it have not aged well.

What still works

The core idea — covering a topic comprehensively across many interconnected pages rather than one long page — is more important now than it was in 2019, not less. AI Overviews specifically favor sites with demonstrated topical depth. A WordPress blog about cooking will not be cited about Core Web Vitals no matter how good the post is. Topic clusters are how you build the kind of depth that earns citation.

The internal linking pattern still works. Cluster pages linking to the pillar, pillar linking back to clusters, related clusters linking to each other. This is how PageRank flows through your site and how Google understands which pages on your site are about which topic.

The keyword strategy still works at a high level. Pillars target broad head terms; clusters target long-tail variants. The pillar accumulates link equity from clusters and external links, often outranking purpose-built money pages over time.

What has aged badly

The "build a 5,000-word pillar page" template. Pillar pages worked when they were genuinely the most comprehensive resource on a topic. In 2026, when AI Overviews can synthesize a more comprehensive answer than any single page can write, the value of being the "ultimate guide" has dropped sharply. A focused 1,500-word pillar that links well to a strong cluster typically outperforms a 5,000-word pillar that tries to cover everything itself.

The exact-match anchor text strategy. Old-school cluster building used exact-match keywords as anchor text everywhere ("best email marketing software" repeated 40 times across the cluster). Google now demotes pages with unnatural anchor patterns. Use varied, natural anchors that match how a writer would actually phrase the link.

The hub-and-spoke linking with no horizontal links. Original cluster diagrams showed pillars in the center with spokes radiating out to clusters. The clusters didn't link to each other. That pattern produces brittle topical signals — clusters benefit from cross-referencing each other, not just the pillar.

The 50-cluster ambitions. Most small businesses planned huge cluster maps with 30-50 supporting pages per topic. They shipped 5 and abandoned the rest. A small cluster done well outranks a half-built large cluster every time. Plan for what you'll actually finish, not what you'll outline.

What to do instead in 2026

Build smaller, denser clusters. Eight to fifteen tightly related pages with high cross-linking density. Pillar at around 1,500 words. Each cluster page at 600-1,200 words. Every page contains 3-5 links to other pages in the cluster, not just to the pillar.

Match cluster topics to AI search query patterns. AI engines respond to question-shaped queries — "how does X work," "why is X important," "what's the difference between X and Y." Build clusters around question taxonomies, not just keyword lists. Each cluster page answers one question completely; the pillar synthesizes them.

Treat the cluster as a unit for E-E-A-T purposes. One cluster, one author byline. The same expert covering the related topics builds authority faster than a different author per page. If you have multiple authors, segment them by cluster, not by article.

Add structured data at the cluster level. The pillar gets Article schema. The cluster pages get Article + BreadcrumbList that places them in the cluster. The combined graph signals to Google and AI engines that these pages are a unit covering a topic, not isolated posts.

The diagnostic

If you have an existing cluster strategy, three quick checks tell you whether it's working:

1. Is the pillar ranking for its target head term? If yes, the cluster is doing its job. If no after 12 months, the cluster is too small or the cross-linking is too weak.
2. Are individual cluster pages getting impressions for long-tail variants you didn't explicitly target? If yes, your topical authority is being recognized. If only the pages you specifically optimized rank, you're getting credit per page rather than per cluster.
3. When you publish a new page in the cluster, does it rank faster than a similar new page outside the cluster? This is the cleanest test of whether the cluster is providing topical lift. If new posts inside the cluster have a meaningful head start over standalone posts, the cluster is functioning.

If you fail one or more of these tests, the most likely fixes are: tighten internal linking density (every cluster page should link to 3-5 others), trim or merge overlapping cluster pages, and rewrite the pillar to actually be the best entry point to the topic rather than a sprawling reference.

If you want to map your existing internal linking structure, run a few key pages through the AI Search Readiness Checker and look at the internal link analysis section. Pages with too few internal links from the rest of the cluster are easy to spot — they're the ones quietly underperforming despite being well-written.