Does Topical Authority Matter for AI Citations?

I had a hypothesis: AI systems prefer sites that demonstrate expertise through multiple related articles. The data seemed to support it. But correlation isn't causation, and I wanted to know if building content clusters actually caused better citation rates.

So I ran an experiment.

The test

I took three sites in different niches, each with a single article on their target topic. Good articles, well-optimized, but isolated. Baseline citation rate across the three: about 2%.

Over six months, I built content clusters around each article. Five supporting articles per month, all interlinked, all targeting related subtopics.

By month three, citation rates had doubled to around 4%.

By month six, they averaged 7.3%.

The control pages I was tracking on the same sites, unrelated content without clusters, stayed flat.

What seems to work

Pillar plus subtopics. The structure that performed best: one comprehensive pillar article (2,500+ words) surrounded by 5-10 focused subtopic articles (1,200-1,500 words each).

Each subtopic linked to the pillar and to 2-3 related subtopics. The pillar linked out to all the subtopics.

Consistent publishing beats batches. I tried both approaches. Publishing one article per week for 10 weeks outperformed publishing 10 articles at once. I think the steady updates signal active expertise.

Genuine subtopics, not padding. Each supporting article needed to answer a real question. I tried creating articles just to add to the cluster, and those didn't help. The cluster only worked when every piece actually earned its existence.

The timeline

This isn't fast.

Month 1-2: Foundation. Pillar article plus 2-3 subtopics. Citation rates unchanged.

Month 3-4: Traction. 5-7 total articles in the cluster. First noticeable improvement in citations.

Month 5-6: Authority. 8-12 articles. Citation rates stabilized at 3-4x the starting point.

Month 7+: Compounding. Continued adding 1-2 articles per month. Each new article improved the whole cluster's performance slightly.

What I'm still uncertain about

I can't isolate topical authority from content volume. Maybe the improvement came from just having more pages to get cited, not from the authority signal.

I also don't know the minimum viable cluster. Is 5 articles enough? Do you need 10? The data suggests more is better, but I can't pinpoint a threshold.

And I'm unsure how much the internal linking structure matters versus just having the content exist. I did it one way and it worked, but I didn't test alternative structures.

How to apply this

If you're starting from zero on a topic, plan for at least six months of building.

Map out your cluster before you start. One pillar page covering the broad topic, 8-12 subtopic pages each covering a specific question within that topic.

Publish consistently. One article per week is sustainable for most teams and seems to perform better than burst publishing.

Link everything. Every subtopic links to the pillar. Every subtopic links to 2-3 related subtopics. The pillar links out to everything.

Update the pillar when you add new subtopics. This keeps it fresh and adds new internal links.

Measure at the cluster level. Individual page performance will vary, but the cluster's aggregate citation rate is what matters.

It's a bigger commitment than optimizing individual pages, but the results are more durable. Single well-optimized pages get copied. A comprehensive content cluster is harder to replicate.