I’ve been testing a few different content workflows over the last couple of months, and one pattern keeps showing up that I didn’t fully expect.

The smaller, tightly connected clusters seem to outperform the larger “publish as much as possible” approach, at least for the kinds of niche sites I’m building.

What I mean by that:
– 8–15 articles built around one intent/topic group
– Strong internal linking between supporting pages
– One clear money page or lead page
– Content optimized more for usefulness and coverage than raw word count
– A mix of AI-assisted drafting + manual editing + SERP validation

On paper, the bigger approach should win because you’re covering more ground faster. But in practice, I’m seeing the smaller clusters index cleaner, rank a bit faster, and hold positions more consistently. I suspect part of it is topical clarity and part of it is simply better internal linking discipline.

A few observations from my tests:
1. Pages in smaller clusters seem to get crawled together more reliably.
2. Internal links actually matter when the topic set is narrow enough to reinforce relevance.
3. AI content quality is less of an issue when the cluster has a clear structure and not just random keyword coverage.
4. The “publish 100 articles and hope” model creates a lot of thin supporting pages that never really find a role.

I’m not saying large-scale content production is dead. If anything, I still think volume matters in competitive niches. But I’m starting to think most people overestimate how much value they get from broad AI content scaling without a strong site architecture.

Curious if others are seeing the same thing.

A few questions:
– Are you building smaller topical clusters or broader sitewide coverage?
– Have you noticed any difference in indexing speed or ranking stability?
– Do you think internal linking is becoming more important as AI content becomes more common?
– For affiliate sites, are you finding better conversion from focused clusters versus larger content libraries?

My current leaning is that a well-structured 12-page cluster with solid internal linking beats a sloppy 60-page content dump more often than not.

Would be interested to hear what others are seeing in real projects, especially if you’ve tested both approaches on similar domains.