Internal linking: a system, not a habit

Internal links are the only significant ranking input you control completely. Nobody has to agree to give you one. There is no outreach, no budget, no waiting.

They are also, on most sites, the thing done least deliberately — added when someone remembers, pointing wherever felt natural at the time, with anchor text like "read more". A site with two hundred pages and no linking policy has thrown away the one lever it fully owns.

This is a sub-pillar of SEO basics. It is about building a system rather than acquiring a habit.

What internal links actually do

Three distinct jobs, and confusing them is why linking advice is muddled.

They make pages findable. A page linked from nothing is reachable only through a sitemap. Links are the primary route by which pages are discovered and re-crawled.

They describe what a page is about. The anchor text and the surrounding context tell Google what the destination covers. Twenty pages linking to one page with descriptive anchors is a strong, consistent statement.

They distribute standing around your site. Pages that have accumulated some authority pass a share of it through their links. This is why a link from your best page is worth more than a link from a page nobody visits.

Where links should point

The rule most sites get backwards: link from strong pages to pages that need help, not the reverse.

The instinct is to link from every new post to the homepage and the main service page. Those pages are usually already the best-linked on the site. The pages that need links are the ones you want to rank that nothing points at.

A practical priority order:

| Priority | Link target | Why | |---|---|---| | 1 | Pages ranking 5–20 for something real | Closest to a return; a push may be enough | | 2 | New pages, from established ones | Discovery and initial context | | 3 | Cluster pillars, from every spoke | Establishes the hub | | 4 | Commercial pages, where genuinely relevant | Where the traffic is meant to go | | 5 | The homepage | Already linked from everywhere |

Anchor text

Google's guidance is for anchor text that describes the destination. That is the whole standard, and it rules out both common failures.

Too vague — "click here", "read more", "this article", "here". These describe nothing, and they are also poor for anyone using a screen reader, which is a better reason.

Too repetitive — the identical exact-match phrase in every link to a page. This is a pattern rather than an optimisation, and stuffing a keyword into every anchor is the kind of thing Google's guidance warns against.

The version that works is unremarkable: describe what the reader will find, varying naturally because sentences vary.

…which is why reading the results page before you write saves more time than any tool.

…the same problem appears when two pages compete for one query.

Different anchors, same destination-describing job.

Where the link should sit

Position matters more than people expect.

In the body, inside a sentence that gives it a reason to exist. A link a reader would plausibly follow because they want more on exactly that point.

Not in a "related posts" block. Automated related-post widgets are weak signals and readers largely ignore them. They are not harmful; they are just not the work.

Not in a list at the end. Dumping six links under "further reading" is the shape of an article that did not integrate them.

Early beats late, mildly, because more readers reach it.

The system

Four rules, applied every time something is published. That is what makes it a system rather than an intention.

1. Every new page links up

To its cluster pillar, once, in context, near the top where the reader is orienting.

2. Every new page links across, two or three times

To siblings covering genuinely adjacent questions. Not to fill a quota — where a reader would actually want them.

3. Every new page gets two inbound links from existing pages

The step everyone skips, and the one that matters most. Publishing a page and moving on leaves it linked from nothing but the index. Before you consider a page finished, go to two existing pages and add a contextual link to it.

This is the difference between a library and a pile.

4. Every new page links to a commercial page, where honest

If the subject genuinely connects to something you sell, link it once. If it does not, do not manufacture a reason — a forced commercial link reads exactly as forced.

Finding the pages that most need a link

The priority table above is only useful if you can identify the pages in row one, so here is the concrete version.

In Search Console, open Performance, set the date range to the last three months, and add a filter for average position between 5 and 20. That list is every query where Google already shows you and people already do not click.

Now take the landing pages behind those queries and ask, for each: how many pages on my site link to this? Anything with fewer than three is a page that is already nearly ranking and is being starved by your own structure. Adding two contextual links from your better-established pages costs half an hour and is the highest-return half hour available on most sites.

Do it in small numbers — five or six pages, then wait a month and look again. Changing two hundred links at once tells you nothing about which change did anything.

Auditing what you already have

Three checks, worth an hour a quarter.

Find orphans. Pages linked from nothing but navigation or the blog index. These are invisible in the structure even though they exist. Crawl your own site or check which pages appear in no other page's links.

Find over-linked pages. If a quarter of your internal links point at the homepage, they are doing nothing useful. Redistribute.

Find the near-misses with no links. Cross-reference pages ranking 5–20 against pages with fewer than three inbound internal links. That intersection is the highest-return list on your whole site.

We run this check on this site with the crawler that gates publishing: it errors on any indexable page nothing links to, and warns when a page has only one inbound link. Automating the check is what stops it depending on someone remembering.

Two things that quietly break internal links

Both are common, both are invisible unless you look, and both waste links you already have.

Redirect chains. You change a URL, set a redirect, then change it again a year later and set another. Now every internal link pointing at the original passes through two hops. Nothing is broken, exactly, but you are relying on a chain for links you could simply repoint. When you change a URL, update the internal links to it as well as adding the redirect — the redirect is for the outside world, which you cannot edit.

Links to pages that no longer exist. Deleted a post, merged two pages, retired a service. The links from your other pages remain, pointing at a 404. Readers hit a dead end and the link does nothing. A crawl of your own site finds these in minutes; the fix is usually repointing to whatever replaced it.

Neither of these is exotic. Both accumulate quietly on any site older than a couple of years.

Ecommerce and larger sites

The rules above hold, with one addition that matters at scale.

On a site with thousands of URLs, most internal links are generated rather than written — navigation, category listings, filters, related-product modules, pagination. Faceted navigation in particular can generate an enormous number of near-identical URLs, each one absorbing links, none of them worth ranking.

The useful discipline is to be deliberate about which generated links exist at all, and then to add the handful of *written* links that generated ones cannot produce: a genuine contextual link from a buying guide to the specific category it discusses is worth more than a hundred automatic ones, because it says something a template cannot.

What internal links cannot do

They redistribute what you have. They do not create it.

A site with no external standing can link its pages together perfectly and still not compete for contested terms — there is nothing to distribute. Internal linking makes the most of your position; it does not change your position. That is what backlinks do, slowly.

They also cannot rescue a page that answers the wrong question. Links help a page be found and understood; they do not make it the better answer.

If the honest reading is that your subject's queries are held by sites far stronger than yours, no amount of internal structure changes that this year. The realistic options are narrower targets or paying for position directly — when renting attention beats owning it sets out the trade, and the rankings board shows the current prices.

Common questions

How many internal links per page? As many as genuinely help. On a 2,000-word article that is usually three to eight.

Do links in navigation count? They help discovery. They say much less about any individual page, because they are on every page.

Should I nofollow internal links to control flow? No. That was a tactic long ago and it does not work as people imagine.

Is it too late for an old site? No. Auditing an existing library usually produces faster gains than publishing, because the pages already exist and are already understood.