Is Internal Linking the Most Underrated SEO Trick?

My Medium friends can read this over on Medium.

A thread I read recently opens like this: "internal linking is the most underrated SEO trick you can pull, and almost nobody does it." The reason: "because it's boring."

So the author has Claude do it. Every week. Prompt included.

I like internal links. I used the line myself in 2022, that few site owners know what their link structure does.

But every week? So I had Claude Code run the first two steps of that prompt on my own data. Three months of Search Console numbers, my whole archive. No edits, only the table.

That prompt

The short version:

• Export three months of Search Console data and connect your site to Claude.
• For each article, take the queries where it ranks on position 1 to 10. Those are its anchors.
• Go through every post. Where it mentions another article's query, turn that phrase into a link. Max 3 to 5 new links per post, with priority for pages on position 4 to 15.
• Show a table first. Edit only after your OK.
• Repeat every week.

The table first is a good rule. Remember that one.

Evidence

Internal links do help. That’s true.

seoClarity ran a split test on roughly 7,500 URLs for four weeks. The linked pages got 1.3 percent more clicks, the control pages lost 17.9 percent. Those were location pages, and seoClarity sells the testing tool. But it's a controlled test.

Zyppy looked at 23 million internal links across 1,800 sites. URLs with 40 to 44 internal links got four times the Google clicks of URLs with 0 to 4. After about 45 to 50, traffic goes down again. The author calls it mostly a correlation study.

The test closest to a blog is an older one from SearchPilot. An ecommerce content hub went from two to four related article links per page. The pages giving the links gained 16 percent. The pages receiving them: "no detectable impact."

And the prompt is all about the pages receiving them. To be fair, those pages had a lot less traffic, so the test may have been too small to see a change.

So, helpful. But "most underrated"? I bet every SEO guide you've seen has a section on internal links. Mine does. And backlinks are still the big chunk of SEO. The thread says that too: "still build backlinks."

My table

Substack first. Roughly 880 posts. I set "decent impressions" to 50 in three months, and the script found 320 possible links.

310 of them use the same word: "notion."

Two of my pieces rank for it. So both become targets, from about 155 posts each. In roughly 95 of those posts, it's a "Notion template" or "Notion-based" line, many of them old ads of mine: "Get my Notion template, Moviebuff, 50% off today."

And six are just the word. "…reinforces the notion that Medium should be viewed as…" Claude would catch those. I hope.

The better of those two pieces got 5,322 impressions for "notion" in three months, at position 7.8. Clicks: zero.

People who type "notion" into Google want Notion. Not my take on it.

317 of the 320 links would use an anchor that got no click in those three months. Most of that is "notion", of course. But of the 109 anchors step one found on my Substack, 73 got zero clicks. On Medium, 377 of 806.

The prompt filters on impressions. So a query where Google shows me and nobody clicks counts the same as a query that brings readers.

Medium looks better. 85 possible links, with anchors like "medium or substack", "best blogging platform", "icloud". Some of those I would take. But 52 of the 85 had zero clicks too.

Week two

The thread says to repeat this every week.

No post in my archive had more than three candidates. So the limit of 3 to 5 never kicks in. The first run takes everything. The second run is an empty table.

With 10 impressions instead of 50, Medium lasts two runs. Substack still one.

After that, only new posts have anything to link. One article at a time. "Internal links between old and new" is on my checklist at publish time anyway.

Position 4 to 20

"the pages sitting on position 4-20 are the ones that move the most."

I checked that too, no prompt involved. May to July against August to October, all my queries with at least 50 impressions in both. The median move:

• Position 1 to 3: 0.85
• Position 4 to 10: 1.15
• Position 11 to 20: 4.15
• Position 21 to 50: 6

So yes, those pages move. Pages further down move more. Without any weekly run.

And moving doesn't mean moving up. On Medium, 57 percent of my page-two queries went up. On Substack, 92 percent went down, "blogging platforms" from 12.6 to 34.6. (That's 44 and 37 queries, and 27 of the Substack ones are about blogging. So mostly one topic falling off.)

If I had started this prompt on Medium in July, I would have a great thread now.

Same with "leave the site alone for a few days." Rankings can move within days. I wrote that myself about adding internal links. I had no control group either.

There's also a small bug. Step one only accepts queries on position 1 to 10 as anchors. The rule wants to push pages on 4 to 15. The thread's closing line says 4 to 20. So "substack seo", where the Medium version of the story I just linked sits at 12.1, never becomes an anchor. On Substack, 9 of the 40 posts with a query in that range get no anchor at all. On Medium, 8 of 85.

There is always a catch

My test is a script, not a full Claude run. It only counts exact phrases. The prompt allows close variants, so Claude could find more. And Claude would probably throw out many of those "notion" rows. That's what the table is for.

It's also my site. Two platforms, a long archive, a lot of old digests. A small company site with 40 well-kept pages will look very different.

And I didn't make any of these edits. So I can't tell you what those links would do. Only what the table looks like.

My version

If I run this, it's once. With these changes:

Attached: my Search Console export for the last 3 months
(query + page).

1. For each article, collect the queries on position 1-20 with
   at least 1 click. Those are its anchors.
2. Find posts on a related topic that mention one of those
   anchors in the body. A mention alone isn't enough.
3. Rules: max 3 added links per post in total, max 10 new links
   per target. No headings, intro sentences, existing links or
   self-links.
4. Split the targets in half at random. Only link to one half.
   The other half is my control group.
5. Show me a table first: source post, anchor, target, query,
   position, clicks. Wait for my OK.
6. After my OK, make the edits and give me a changelog so I can
   revert.

Then I would look again after four to eight weeks. New posts get their links when they go out.

I haven't run this version yet. So no promises.

The Bottom Line

So, is internal linking the most underrated SEO trick? I don't think so. It's a basic. A useful one. And yes, boring. The thread got that right.

Claude can do it. Once. I would read that table before saying OK.

Mine had "notion" in it 310 times. What's in yours?

If you're not already, follow me on Medium.

Get Substack SEO