New Internal Linking Tool in Data Wise 🔗

Hi all.

I wanted to show you a cool new tool that we've integrated into data wise, the internal linking tool.

This is going to give you all the possible internal links that this tool has found for you with the given website.

For those of you who remember last week I gave you guys a jev internal linking GitHub repo, which you can access now, essentially, now we've turned that GitHub repo into a tool.

I want to make sure that we understand how this application works before you kind of use it.

I think it's really, really important that you know the full kind of backend perspective of it.

So here you can see that it saves all the past runs.

I've done a couple of websites here.

and what it tells you is, you know, from the website that we gave it, it found 12, opportunities or 12 backlinks that it knows that you should probably do or, you know, it's very, very sure you should do.

There's 16 that needs your review that it's kind of semi sure.

And there's some that is really not sure that you should internally link.

You can download this as a CSV, give it to someone to do, or if you've built your website with Claude and Astro, you can just give that CSV to Claude and say, hey, these are the internal linking opportunities.

Anything that's been, classified as sure, please internally link it.

I want to make sure you know how this whole thing works.

I think that's really important, so that you know the backend parameters and maybe you don't agree with the way that I've approached this, which is more than enough, fair enough.

But, I think it's going to work pretty well.

So there are four steps that are happening here.

We are crawling, we're reading the site map and pulling every paragraph that could hold a link.

We're matching, we're finding the four pages that are most related to that paragraph.

Jev is designing meaning Jeff picks the page, or none, depending on whether it's good enough.

And then we place, we pick the exact words in your paragraph that become that hyperlink.

You see here the models we've chosen for that.

the three models in question.

C is BGE M3.

This is an embedding model that measures how close a paragraph and page are in meaning to one another.

we're obviously using JEB from Typescript and then we're using Deep Seq v4 for any other reasoning.

This is using your API key, that you've placed into Data wise, but it is, you know, extremely cost effective.

It's very rare that even for a website over a hundred dollars, over a hundred pages, sorry, you're going to use more than $1.

I want to make sure that you understand the rules that are made here.

So what happens is that each paragraph on the page that gets crawled gets shortlisted to four pages that it could potentially link to.

Right.

And I want to show you the questions that JEV is asking itself.

So once it reads a paragraph in a page, it short lists four pages that it could link to and these are the five questions that JEV is asking itself.

I don't think one question is good enough.

I think there are several components that attribute to a good backlinking strategy and thus the five questions are the best target.

So which pages is the most useful next step for this reader out of the four that we've shortlisted?

Link warranted.

Does this paragraph raise a topic a reader would want to explore?

Is it a good flow for them to go on, anchor available?

Is there a phrase that could become the link without a rewrite so we don't change much, which is internally linking to an existing word reader stage?

Would they be, if they go on from this paragraph to that other page, would they be learning, comparing or ready to act?

And if it's commercial, is there a paragraph, promotional rather than educational?

Right.

All links must clear all those questions to be, appropriately selected.

We give it a weight for each one of those.

The links are spread out, never stacked, so new links per page at most, there's only eight new links to any one of the pages.

And, there's a 75% value of each extra link to the same page and two paragraphs between the links.

At least you're going to get these three different outputs where it's sure where JEV is 88% confident that the anchor link matches the target topic needs review.

It's between 70 and 88%.

you know that it should be internally linking and it's not sure, it's just 50, 70% or above or below.

Sorry again.

It runs on your open router API key.

I've done 30 pages and a website with 300 pages, which is the max, by the way, and I've only wasted, or used, not wasted, 30 cents.

So even if you're running this on a very large website, you really won't use that much credit from open router.

So back on, data wise, you go to the internal link section.

You're going to go here I'M going to go www.datawise SEO.com just so I can show you.

And I'm going to find, link opportunities.

It's reading the site map now.

Again, like I said in the beginning, depending on how your website is, hosted, it might be blocking this.

If it is, please let me know and then we can work on the next steps on how to unblock certain, hosting components and stuff like that.

But it'll tell you.

You can see here, it's crawling the 30 pages.

This site has 418 paragraphs.

It's short listing the candidates pages for each paragraph, yada yada.

This should take about five minutes.

So with some editing magic, I'm going to come back when that is done.

Okay.

Now after a couple of minutes, we see the Jeb ran 417 paragraphs, of course, 30 pages.

And, it costs seven cents.

You get a little, pie graph.

Graph here telling you, you know, the percentage.

Like out of all that we had, reviewed, there's a 19, there's 19% of them that we're really sure you should internally link.

you can leave this here and just go.

Cool, I've done that.

Internal link.

Boom.

I just wanted to do like a quick checklist for you guys and it should save again if you go here.

I need to fix that.

But we'll make sure that once you've clicked I've added it, then it'll save it.

you're better off in exporting that and putting it somewhere if you're sharing it with your team, whatever that is.

But, I hope that's useful.

Let me know if you like it.

And yeah, cheers.