What this check inspects
Shipwork reuses one crawl of your pages, builds the internal link graph from it exactly as the link graph check does, and finds the pages with few or no inbound links from the pages it read. For each one it ranks the other readable pages by how close their text is and picks the best sources, then writes the anchor text from the title of the target page, or its most distinctive description words when the title is unusable. It flattens every suggestion into a single ranked plan ordered by the value of one edit: a target that the near-miss check proves already has search demand comes first, then the least-linked target, then the closest source. So the first line is the best next link to add anywhere on the site, not just the best source for the first page.
What a failure means
This reuses a sampled crawl, so it says plainly what it examined. When it did not read the whole sitemap it does not claim a page is an orphan site-wide, only that it had few inbound links among the pages read. With too few readable pages or no text model, it reports a reason rather than a recommendation. A page that needs links is a warning when the crawl was confident, and an informational note when it was not. A page that needs links but has no close enough source, or no usable anchor, is reported separately so it is not silently dropped. When every page examined is already well linked, it says so.
How to fix it
- Add the ranked links one at a time, starting at the top. That order already weighs a target search demand and how few links it has.
- Use the anchor text the plan supplies, taken from the title of the target page, rather than click here or a bare URL.
- Link from the page named as the source, since it was chosen for being close in topic to the target.
- Where the plan could not write an anchor, write a short phrase that describes the target page and link with it.
- Re-crawl after the edits. The plan is based on the links the crawler could see.
A typical failure, worked through
The setupA store has a size guide and a returns page that almost nothing links to, while the same crawl shows a collection page already appearing for a query just off page one.
What the check reportsThe plan lists the underlinked pages, writes an anchor for each from its title, and puts the collection page first because the near-miss check shows it has proven demand.
The pointThe output is a single edit a person can make, in the order that will do the most good first, rather than a report that ninety-eight pages have no inbound links.