Has anyone actually seen consistent gains from AI-assisted internal linking at scale?

[mercer]

I’ve been testing a few AI-assisted internal linking workflows on one of my smaller niche sites over the last couple of months, and I’m seeing mixed results so far.

The setup is pretty simple:

– Site is on WordPress
– Roughly 180 indexed posts
– Mostly informational content with a few affiliate pages
– I used an AI workflow to suggest internal link opportunities based on topical relevance
– Then I manually reviewed the suggestions before publishing updates

What I’m noticing:

1. **The obvious links are obvious**
The AI is decent at finding generic related-post connections, but those are usually the same ones I’d have added manually anyway.

2. **It sometimes overlinks weak matches**
It tends to push links where the semantic similarity is there, but the user intent is not really aligned. I’m being careful not to create a mess of “related” links that don’t actually help navigation.

3. **A few pages did improve**
I’ve seen some movement on mid-tier pages that were sitting just outside the top 10. Nothing dramatic, but enough to make me think internal linking is still one of the more underrated levers when the site already has decent content.

4. **Scale is the real question**
On a small site, manual review is manageable. On a larger content library, I can see the appeal of automation, but I’m not convinced fully automated internal linking is worth the risk unless the system is very tightly controlled.

My current view is that AI can speed up the process, but it doesn’t replace judgment. The best use case might be:
– cluster-level link suggestions
– anchor text variation ideas
– identifying orphaned or underlinked pages
– updating old posts with contextual links during refreshes

I’m curious how others are handling this.

Has anyone here run AI-assisted internal linking across a larger site and actually seen consistent ranking or crawl improvements?

Also interested in whether you’re using:
– pure AI suggestions
– plugin-based automation
– custom scripts / embeddings
– manual review only
– some hybrid workflow

My suspicion is that most of the value comes from better site architecture and content grouping, not the AI itself. But if someone has data showing otherwise, I’d be happy to be wrong.

[Author]

Posts