I’ve been testing a few different internal linking workflows across small niche sites over the last couple of months, and I’m curious whether others are seeing anything similar.

My setup is pretty simple:

– WordPress sites in the 20–80 article range
– Mostly informational content with a few affiliate pages mixed in
– Manual internal links added during publishing, plus a second pass using an AI-assisted workflow to suggest contextual links
– A few sites also use topical clusters and “related articles” blocks

What I’m noticing is that internal linking still seems to be one of the highest-ROI things you can control, but the results are not always as dramatic as some tools make it sound.

A few observations so far:

1. **Pages with stronger internal support do seem to move faster**
Especially newer pages that would otherwise sit in the middle of the pack. I’ve seen some jump from nowhere to page 2/low page 1 more quickly after adding links from relevant older posts.

2. **Context matters more than raw link count**
I tested adding a bunch of links sitewide versus fewer links placed naturally in the body. The sitewide-style approach looked cleaner from an automation standpoint, but the contextual links performed better in terms of engagement and, in a few cases, rankings.

3. **AI suggestions are useful, but not reliable enough to run blindly**
The models are decent at finding semantically related pages, but they still miss intent mismatches. I’ve had it recommend linking from a broad “best X” page to something too narrow, or from an informational page to a commercial page where the anchor felt forced.

4. **The biggest lift may be from fixing orphaned or underlinked pages**
This is probably the least glamorous part, but it’s where I’ve seen the clearest improvement. A page with 0–1 internal links often benefits more than a page that already has decent structural support.

I’m still not convinced there’s a magic AI internal linking tool that can replace a good site architecture. At best, these tools seem to help scale what you’d ideally do manually anyway.

What I’m trying next:
– Building tighter topic clusters around a few money pages
– Using AI only for link discovery, not final placement
– Tracking whether pages with 3–5 relevant internal links outperform pages with 10+ weaker