Internal Linking for SEO: What Google Documents, and a 106-Page Model of Where Link Equity Goes
· Royking Niba
Internal linking for SEO is the practice of linking your own pages to each other so that search engines can find every page you care about and understand which ones matter most. Google states that it “uses links as a signal when determining the relevancy of pages and to find new pages to crawl”, and that “every page you care about should have a link from at least one other page on your site.” The part most sites get wrong is not how many links they add but where those links start. In the 106-page model below, one link from the homepage lifted a buried article’s share of internal link equity more than twenty contextual links from ordinary articles did.
What Google actually documents
Google’s link best practices page, last updated 10 December 2025, is short and specific. I read it on 3 October 2026. The four lines that matter for internal linking:
Google uses links as a signal when determining the relevancy of pages and to find new pages to crawl.
Google Search Central, Link best practices for Google
Every page you care about should have a link from at least one other page on your site.
Google Search Central, Link best practices for Google
Good anchor text is descriptive, reasonably concise, and relevant to the page that it’s on and to the page it links to.
Google Search Central, Link best practices for Google
On quantity, the same page offers a rule of thumb rather than a number: “if you think it’s too much, then it probably is.” On anchors, it warns you to “resist the urge to cram every keyword” into the link text. Those two lines are the whole of Google’s published position on volume and keyword-rich anchors. There is no internal link quota and no recommended anchor ratio anywhere in the documentation.
What Google does not publish is how much weight internal link structure carries today. So the rest of this page uses a model to show direction, not magnitude.
The model: 106 pages and four ways to link them
I built a small site in code and ran the original PageRank calculation over it. The formula comes from Brin and Page’s 1998 paper describing the first Google prototype, which says “The parameter d is a damping factor which can be set between 0 and 1. We usually set d to 0.85.” I used 0.85. To be clear about the limits: this is the published 1998 formula, not Google’s current system, and it ignores anchor text, link position and everything off the site. It is useful because it makes the arithmetic of link placement visible.
The site has one homepage, five category hubs and 100 articles, 20 per hub. The target is the oldest article in the first hub, the kind of evergreen page that earns money and slowly sinks. Every article links back to its hub and to the homepage in all four versions.
- Version 1: each hub links only to its ten newest articles, and older articles are reached through a “next post” chain, which is how many paginated blogs work in practice.
- Version 2: version 1 plus three contextual links from each article to other articles in the same hub, chosen at random.
- Version 3: version 2 plus twenty extra contextual links pointing at the target from articles across the site.
- Version 4: version 2 plus one link to the target from the homepage, and nothing else.
| Version | Target’s share of internal PageRank | Versus the average article | Rank among 106 pages | Clicks from the homepage |
|---|---|---|---|---|
| 1. Newest ten on the hub, older posts by next-link chain | 0.197% | 0.44x | 102nd | 12 |
| 2. Plus three contextual sibling links per article | 0.461% | 0.70x | 87th | 3 |
| 3. Plus twenty contextual links aimed at the target | 2.103% | 3.16x | 7th | 3 |
| 4. Plus one homepage link instead | 2.259% | 3.35x | 7th | 1 |
What the numbers say
The chain is the problem, not the content. In version 1 the target sits twelve clicks from the homepage and ranks 102nd of 106 pages. Nothing about the article changed between versions; only the links around it did. This is the same mechanism as the 167-page example in pagination SEO: a sequence that only links forward buries everything at the end of it.
Ordinary contextual links fix depth before they fix equity. Three random sibling links per article cut the target’s click depth from twelve to three and more than doubled its share, from 0.197% to 0.461%. It is still below the average article, because random links spread equity evenly. Contextual linking is the floor, not the strategy.
Where a link starts matters more than how many there are. Twenty contextual links from ordinary articles lifted the target to 3.16 times the average article. One link from the homepage did slightly better, 3.35 times, and put it one click from the front door. In version 2 the homepage holds 12.3% of the site’s internal PageRank against 0.66% for the average article, about nineteen times as much, so one link from it is worth roughly twenty from pages that each hold a little. If you want a page to matter, link to it from the pages that already do: the homepage, the main navigation and your strongest hubs.
And the warning that comes with it. Every link added to the homepage dilutes the others it carries. In this model the cost was small because the homepage had only five other links. On a homepage with eighty links, the same single link would carry far less, because the equity is split eighty ways. That is Google’s “if you think it’s too much, then it probably is” in arithmetic form.
Anchor text, in Google’s own examples
Google’s page gives “Click here”, “Read more”, “website” and “article” as anchors to avoid, and “list of cheese types” as a better example. Here is the same test applied to the links I see most often on audits.
| Anchor as found | Problem | Rewrite |
|---|---|---|
| Click here | Says nothing about the destination | how to read a crawl budget report |
| Read more | Generic, repeated on every card | Index bloat: the cost in recrawl days |
| best seo services cheap seo agency seo company | Crammed keywords, the pattern Google warns against | our technical SEO audit |
| this article | Generic | the 12,000-URL orphan page example |
| https://example.com/blog/post-123 | A raw URL tells a reader nothing | the 301 vs 302 decision table |
The useful habit is to write the anchor as a description of what the reader gets on the other side. It is the same skill as writing the one-line notes in an llms.txt file, and for the same reason: something has to decide whether to follow the link based on those few words.
The internal linking audit I run
- Export every indexable URL from the sitemap and every URL reachable by following links from the homepage. The difference between the two lists is your orphan set, which I cover in orphan pages.
- Record click depth for every page. Anything that earns revenue and sits deeper than three clicks goes on the fix list first.
- List the ten pages with the most internal links pointing at them. If they are not the pages you most want to rank, your navigation is voting for the wrong pages.
- Replace “next post” style chains and archive-only access with category hubs that link directly to every article, or to every article in a sub-hub.
- Add three to five contextual links per article to genuinely related pages, with descriptive anchors. This is the floor from version 2 of the model.
- Choose the few pages that matter most and link them from the homepage or main navigation, the version 4 move. Remove a weaker homepage link for each one you add.
- Re-crawl after the change and compare depth and inlink counts. Then check server logs over the following weeks for Googlebot requests to the pages you promoted, using the method in server log file analysis.
Where this sits in a recovery
A site that lost traffic after a migration or a redesign often lost its internal link graph without anyone noticing: a new template drops the related-posts block, a menu shrinks, an archive switches to infinite scroll. Before treating a drop as a penalty, compare the old and new link graphs. The full diagnostic order is in Google penalty recovery, and the depth arithmetic behind this page is in site architecture for SEO.
Leave a Reply