I’ve been testing AI-assisted internal linking across a few niche sites over the last couple of months, and I’m starting to think the results depend a lot more on site structure than on the tool itself.

On paper, it sounds straightforward: use AI to identify contextually relevant anchor opportunities, push links into older posts, and improve crawl paths plus topical depth. In practice, I’m seeing mixed outcomes.

What I’ve tested so far:
– A 120-article affiliate site in the home improvement niche
– A smaller informational site with around 40 posts
– A WordPress setup using a mix of manual review and automation through an AI workflow
– Internal links added mostly to older pages that already had some impressions but weak CTR

A few observations:
1. On the better-structured site, the AI suggestions were actually useful. It found a lot of “obvious in hindsight” links I would have missed manually.
2. On the messier site, it kept recommending links that were technically relevant but felt forced in context.
3. I did see some movement in GSC, but it’s hard to separate link effects from normal content aging and indexing lag.
4. The biggest issue was anchor over-optimization. If you let the system run too loosely, it starts repeating the same exact-match phrasing too often.

My current view is that AI internal linking works best as a support layer, not a fully autonomous system. It seems more effective when:
– you already have solid topical clustering
– pages are mapped by intent
– you limit how many links get inserted per article
– you review anchor diversity manually

I’m curious how others are handling this.

Has anyone here found a workflow that actually scales without making the site look spammy or bloated?
Are you using:
– AI suggestions only
– semi-automated insertion
– fully automated internal linking
– plugin-based approaches in WordPress

Also interested in whether anyone has seen better results from internal linking on:
– affiliate sites
– local SEO sites
– GEO-style content hubs
– informational niche sites

My suspicion is that most of the “AI internal linking” hype is really just good site architecture plus faster execution. But I’m open to being proven wrong if someone has data.