EnterraHost
A link crawl says two different things and only one of them is about links. This is a second look at the same site the first piece measured, with the number that matters most and the number that misleads.
The number that matters
Across the twenty pages the audit's own crawl reads, every page has at least one internal link pointing at it, the median page has 26, and the busiest has 128. Three pages sit at the bottom with exactly one each.
| Measure | Result |
|---|---|
| Pages crawled | 20 |
| Pages with no inbound internal link | 0 |
| Median inbound links | 26 |
| Fewest inbound links | 1 |
| Most inbound links | 128 |
The three pages with a single link in are the author page, the plugin page and one support category. None of them is broken, and all three are reachable. They are one hop from being stranded, because a link can be removed by an edit that has nothing to do with them.
That is the useful measure. A page with one inbound link is a page that depends on one other page continuing to link to it, and nobody is tracking that.
The number that misleads
The longer crawl of the same site read 141 pages, and 137 of them were found through the sitemap rather than by following a link. Only three arrived by following a link from another page, and the homepage arrived as the seed.
It would be easy to publish that as "137 pages nothing links to". It is not true, and the crawl never claimed it was. The crawler is handed the sitemap first, so it queues every address before it follows anything. A page found through the sitemap tells you the sitemap lists it. It tells you nothing about whether a reader can reach it.
The two numbers measure different things and only one of them is about internal linking. The sitemap is how a crawler finds a site. Links are how a reader does. A crawl that reports its discovery depth is reporting its own seeding strategy unless you read it carefully.
What the graph says when the menu is removed
The visualiser separates a link in the body of a page from a link in the menu, and that distinction changes the answer completely. Across the same 141 pages and 3,051 links, the median page has one link pointing at it from the body of another page. The rest of what points at it is the menu.
Five pages have no body link at all. Four of them are reachable through the menu and nothing else, which is why they look fine. The fifth, /authors/network-operations/, has no links pointing at it of any kind. It answers 200, it is in the sitemap, and no reader has ever arrived at it by clicking.
That is the finding the sitemap number hides and the reason to count body links separately. A menu link appears on every page of the site, so it inflates every count while telling you nothing about whether the page is linked to in the sense that matters.
How to get the real answer
Count the links, not the depth. For every page in the crawl, count the links whose target is another page in the crawl, and invert the count. What is left is a list of pages nobody points at, and that list is short, real and actionable.
It also costs nothing to reproduce. The broken link checker returns every link on every page it reads, and the link visualiser draws the graph, which is the fastest way to see a cluster of pages that only link to each other.
What to do with the low end
Do not add links for the sake of the number. A page with one inbound link needs one more link from a page that a reader would plausibly be on when they want it. In this site's case the author page is linked from one byline and belongs in the footer of nothing, while the plugin page should be linked from the support hub that discusses it.
The rule is the same as for outbound links: a link that exists so a crawl will follow it is a link a reader will not.
This site was measured while being written. The about page says who runs it and the legal page carries the company detail, and until this piece both were reachable only from the footer.