One page on my own network has 44 internal links pointing at it. Forty-three of them say “see how it works.” I did not find that by hand. I found it by crawling all 25 of my WordPress properties on 3 September 2026 and counting every link in the body of every page, and it is the single most embarrassing number in the dataset.
Here’s what you actually get when you count your own internal links properly, and the three-step audit that comes out of it.
Seven in ten of your internal links are furniture
The crawl covered 1,350 published posts and pages and found 17,256 in-content links. It also found 41,311 template link instances — nav, breadcrumb, sidebar, footer. That is 70.5% of the whole graph, and it is the same handful of anchors repeated on every URL. seoarcade.com alone repeats 60 template links across 453 pages.
That matters for one practical reason: if you audit internal links with a tool that does not strip site-wide chrome, you are reading your own footer back to yourself. Every page will look well-linked, because every page is linked from every other page by the theme.
Strip the chrome and the real graph is smaller than you think
After removing template links, embedded tweets and navigation, 17,256 links become 13,207 body-prose links, and those resolve to 4,871 body anchors pointing at 1,011 distinct URLs on my own network, from 1,048 source pages. Twenty-five sites. A thousand target pages. That is the actual editorial link graph, and it is the only part of it you control sentence by sentence.
Which brings up the thing everybody asks next: does the wording of those anchors matter? I measured that separately and the answer was no — term overlap between anchor and query correlates with clicks at ρ 0.04, p 0.55. The full method and the trap that nearly produced a false positive are in the anchor text study. What did correlate was the number of distinct sites linking in, at ρ 0.33.
The three numbers worth pulling out of your own crawl
You do not need my dataset to run this. You need a crawl of your own site with template links excluded, and three counts:
- Orphans. Pages with zero in-content links pointing at them. I had 107. Every one is a page the site itself never recommends, which means the only way in is search or luck.
- Wasted anchors. Any target where most inbound anchors are the same string. Mine was 43 of 44 identical. Those 43 links are doing the work of about one.
- Mismatches. Links whose anchor describes something other than what the destination page is about. I had 22 of those, and they are usually a copy-paste of a CTA.
Orphans are the ones to fix first, because the fix is free. You already wrote the page. And if the destination is a page that sells something, the anchor is worth choosing deliberately — the same logic Bradley Benner applies to citations, where a brand mention next to the right words is the signal, applies inside your own site too.
Why “see how it works” is worse than a bad anchor
A bad anchor sends a confusing signal. A repeated CTA anchor sends no signal at all, forty-four times. The page in question sells a service, so every mention of it across the network got the same call to action stapled to it, and the aggregate is a page whose inbound anchor text describes literally nothing about what it does.
Fixing it is not a rewrite. It is opening the twelve highest-traffic source pages and changing the words inside the existing link so they describe the destination. The link count does not change. The link stays where it is. Only the sentence around it moves.
What I can’t tell you
This is a snapshot, not an experiment. I counted what exists and correlated it against Search Console and GA4 for the window 1 May 2025 to 1 September 2026. I did not change 107 orphans and watch what happened, so I cannot promise you a number on the other side of the fix. What I can tell you is that a page nothing links to is a page you decided not to sell, and you probably did not decide that on purpose.
So go pull your crawl, exclude the chrome, and count your orphans. How many of them are pages you would have sworn were part of the site?