Internal linking for machine readers
Internal links do two jobs now: they move a reader along a chain of questions, and they tell a machine which of your pages belong to the same subject. Descriptive anchor text in a sentence does both. A navigation block does neither well.
Internal linking used to be discussed almost entirely in terms of distributing authority. That still happens, and it is now the less interesting half. The more interesting half is that links are the clearest statement you can make about how your own pages relate.
Five patterns that do the work
Where a page raises something it does not settle, link the page that settles it, in the sentence that raised it. This is both the most useful link for a reader and the clearest relationship statement for a machine.
"What a domain migration costs" tells both readers what they will get. "Read more" and "click here" tell nobody anything, and "our services" tells a machine that the destination is about your services rather than about the subject.
Pages that answer adjacent questions should reference each other. A cluster where everything points at the pillar and nothing points across reads as a hierarchy rather than a subject.
To the regulation, the dataset, the standard, the source. Outbound links to authoritative material are a signal about your own carefulness, and their absence on a page full of specifics is conspicuous.
Anything worth retrieving should be within three clicks of the home page. Depth is a crawl budget problem on large sites and a discoverability problem on small ones.
An anchor that says "read more" is a link you have chosen not to explain.
What to stop doing
- Footer blocks listing every page on the site. They are ignored as navigation and they flatten the relationships you were trying to state.
- Automated related-content widgets that pick by recency. They link unrelated pages and produce a subject map nobody intended.
- Linking every occurrence of a term. One link per destination per page, at the most useful moment.
- Linking to a page that redirects. One hop, always, and fix the source rather than relying on the redirect.
- Links that only appear after a script runs. They are invisible to anything that fetches without executing.
A check worth running quarterly
Crawl your own site and list every page with no incoming internal links from the body of another page. On most sites this list is longer than anyone expects, and it is full of pages somebody wrote carefully and then nobody ever pointed at.
Questions this settles
As many as the page genuinely raises questions it does not settle, which is usually three to ten in a knowledge page. Counting them is the wrong instrument.
Yes, and arguably more, because it is a plain statement of what the destination is about, which is useful to a retrieval system as well as a ranking one.
No. Internal nofollow as a sculpting technique has not worked for many years. If a page is not worth linking, do not link it.