Single keyword-targeted pages vs topic clusters — what actually ranks in 2026?
Old playbook: one page per keyword. Newer advice: build topic clusters / pillar pages and let internal linking plus semantic coverage do the work. With AI Overviews eating clicks, which model actually wins now?
I have limited writers. Do I spread them across 30 narrow keyword pages or 5 deep pillar hubs? Looking for people who've tested both, not theory.
Tested both across two sites in the same niche over 18 months. The deep-hub model won decisively, but the nuance matters:
Narrow one-keyword pages cannibalize each other constantly. You end up with five thin pages half-ranking for variations of the same intent, splitting signals. Google has been consolidating intent for years — it wants the single best resource, not five mediocre ones.
The cluster model works because it maps to how Google models topics now: a strong pillar covering the head intent, surrounded by supporting pages for genuinely distinct sub-intents, all interlinked with descriptive anchors. Authority concentrates instead of scattering.
BUT clusters only beat single pages if the sub-pages target genuinely different intents. Don't split 'best running shoes' and 'top running shoes' — that's the old cannibalization trap in a cluster costume.
On AI Overviews: the pages cited in AIOs have clear, extractable, well-structured answers and real authority. Deep hubs with crisp sub-sections get pulled in far more than thin keyword pages. So clusters are also your AIO-citation strategy. With limited writers: 5 deep hubs, every time.
The 'cluster costume' line is gold. We had 'cheap X' and 'affordable X' as separate pages and they fought each other for a year.
Merged 6 thin pages into one 3,000-word hub, redirected the rest. Traffic to the hub exceeded the sum of all 6. Consolidation is real.
Counterpoint: in very long-tail spaces, narrow pages still win because the intents really are distinct. Depends heavily on niche breadth.
@Violet 100% — long-tail with truly distinct intents is the exception. The failure mode is splitting ONE intent into many pages, not covering many real intents.
How do you decide pillar vs sub-page? I struggle to know where to draw the line.
@Theo one SERP = one page. If two queries return basically the same top 10, they're one intent → one page. Different top 10s → separate pages.
AIO citations jumped for us after adding clear H2 question headers + concise 40–60 word answers under each. Structure matters more than length now.
Internal anchor text is the part people skip. Descriptive anchors between cluster pages did more than any external link campaign last year.
Watch out: huge pillar pages can get unwieldy and bury the head answer. Keep the pillar skimmable, push depth to sub-pages.
With AIOs killing clicks, is ranking even worth it? Genuine question.
@Aaron being the cited source in an AIO still drives brand, and the clicks that remain skew to cited sources. Invisibility is worse. Play to be cited.
We track 'AIO citation share' as its own KPI now alongside rank. Changes how you write — for extraction, not just for blue links.
Don't forget to actually interlink the cluster. I built 8 great pages and forgot the links between them. Left huge value on the table.
Pillar + cluster also makes refresh cycles sane. Update the hub quarterly, sub-pages as needed, instead of 30 orphan pages rotting.
Saving. 5 hubs > 30 thin pages, write for AIO extraction, one intent per page. Got it.
18 months of testing in one post. This is why I read this forum instead of LinkedIn takes.