EnterraHost
An internal link graph is a picture of how your pages point at each other. It answers a question no page-by-page check can, because it is about the shape of the site rather than the condition of any one URL.
You can crawl a site and draw it without an account. This is what the picture and the counters are saying.
Why a menu link is not a recommendation
The single most misleading number in any link report is an inbound link count, because every page on a typical site carries the same menu and the same footer. Counted as ordinary links, that gives every entry in the menu the same inbound count, and the graph ends up describing your template rather than your writing.
So the tool counts two kinds of link separately. A body link is one a person wrote inside the content of a page. A menu link is one the template adds to every page. Only the first says anything about how your pages relate to each other, and the counters throughout this guide use the first kind unless they say otherwise.
The distinction matters most for the pages that only the menu links to. Those pages are reachable, so nothing is broken, and they have no inbound recommendation at all. Reachable and recommended are different states, which is the whole point.
A body link arrives with a reason attached, which is why a reader follows it and why a menu item gets skipped. That is the benefit to the person, and it is the whole of the search benefit as well, because an internal link is how a crawler works out what a page is about and how much the site values it. A page that nothing links to is hard to rank however good it is, and no amount of keyword work makes up for it. Writing the link where it belongs serves both at once, and it is the only part of this that needs a person rather than a template.
What an internal link graph reports
Pages is how much of the site the crawl saw, and it is the number that qualifies every other one. A crawl capped at a fixed page count sees a portion of a large site, and the shape it draws is the shape of that portion.
Body links is the total of the links a person wrote. Average body links in is the more useful figure, because a total spread across a site can hide a page with nothing pointing at it.
Deepest hop is the number of clicks from the homepage to the furthest page the crawl reached, and pages three or more hops is how many sit out there. Both describe how much of your site depends on the menu to be found at all.
Orphan pages, and the two kinds
An orphan is a page nothing in your content links to. The tool splits them, and the split is the useful part.
Navigation reachable orphans can be got to through the menu, so a visitor who looks will find them. They are in the sitemap and they carry no noindex tag, which means the site is asking for them to be found while no page recommends them. The crawl of this site found two.
No route in orphans have no path to them at all. A page in that state is live and invisible, and unless something external links to it, the only way anyone arrives is by typing the address.
A page that is noindex, or deliberately kept out of the sitemap, is not listed even when nothing links to it. That is correct behavior rather than a gap, because the site has said it does not want the page found.
Click depth is a usability guideline, not a ranking rule
The three-click rule gets repeated as though a search engine enforces it. It does not. There is no documented click depth limit, and the idea is a usability guideline that predates anyone worrying about crawlers.
Depth still matters in practice, for a reason that has nothing to do with a penalty. A page four hops from the homepage receives fewer internal links, is reached less often, and is discovered later when it changes. The guidance on crawlable links is about making sure a crawler can follow what you wrote, and a page nobody links to is the extreme case of that.
The practical fix is dull and effective. Link to the useful pages from inside the pages that relate to them, and the depth problem disappears as a side effect.
What a graph cannot tell you
It cannot tell you whether a link is any good. A link from a page with nothing in common with the target is a link, and the graph draws it the same way as a link from a page that describes it well.
It cannot tell you what the anchor text says, which is the part a person editing the site can improve. And it cannot weigh a link. A menu link appears on every page, which makes it look like the most connected thing on the site, and it is the least informative link there is.
What it does well is show you a site you did not write. Run it on a site you have inherited and the shape appears in a few seconds, which is usually faster than reading the navigation by hand.
Reading your own graph
The crawl above is this site. It reports two navigation-reachable orphans, and they are the pages a visitor is least likely to stumble into, because nothing recommends them to anyone. The counters say the site is reachable and the graph says which parts are only reachable.
That is the difference between a check and a picture. The broken link crawler tells you whether a link works. This tells you what the collection of them adds up to, and the answer is sometimes that a page exists, works, and is effectively unlinked.