Internal linking is the cheapest SEO you're not doing

Every company blog we take over has the same defect. The posts are decent, the archive is respectable, and almost nothing links to anything else. Internal linking is the cheapest ranking work available — no budget, no new content, no permission from anyone — and it is close to universally neglected. The reason is not laziness. It is memory.

Links are how the rest of your site gets found

Google's documentation is unusually blunt here. It finds pages by following links from pages it already knows, and those links must be real anchor elements it can crawl. Google's own guidance on crawlable links spells out both halves: the link must be followable, and its text should describe where it goes.

There is a consequence people skip past. A post that nothing links to depends entirely on your sitemap for its existence. It can still be indexed. It just has nothing telling Google — or a reader — that it belongs to a body of work rather than sitting alone.

The SEO starter guide makes the same point from the reader's side, which is the honest one. Links are how someone who arrived for one answer finds out you have twelve more.

The problem is memory, not effort

Consider what placing one good link actually demands. You have to know the destination exists, remember roughly what it argues, judge that this specific sentence is where a reader would want it, and be right that it still says what you remember.

A freelancer writing your ninth post this quarter has read maybe three of your existing eighty. They link to the two posts they happen to remember, which are usually the two most recent. Every link points backwards a few weeks and no further.

Worse, links only ever get added in one direction. The new post links to the old one. The old one — published, approved, closed — never gets updated to point forward. So your best-performing pages, the ones that have been accumulating authority for two years, are precisely the ones that link to nothing recent.

This is a whole-archive problem, and people are bad at those for the same reason they are bad at remembering everyone's birthday. It isn't difficult. It's just too much to hold at once.

An engine reads the whole archive every time

This is the one corner of content work where automation has a boring, unfair advantage. An engine writing post eighty-one has the other eighty available — not as a memory but as a corpus. It notices that the paragraph it just wrote about approval workflow is the argument it made fourteen months ago from another angle, and links there.

It can also do the part nobody does, which is go back. When we publish, the interesting change is often not in the new post at all, but in the three older posts that now have somewhere better to point. That backward pass is tedious, unrewarded, and never survives a busy quarter. It is also the half that compounds — consistency beats bursts applies to link structure at least as much as to publishing.

Anchor text is a description, not a slot

"Read more", "click here" and "this post" waste the one place where words are most legible to both a reader and a crawler. The test is simple: read the link text alone and see whether you know what is on the other end.

The overcorrection is worse than the original sin. Teams discover that anchor text matters and start forcing the same exact keyword into every link, which is how an archive ends up with forty links reading "AI content writing service" all pointing at one page. Google's spam policies are aimed chiefly at links built to manipulate ranking, and internal navigation is not mainly what that guidance addresses — but the instinct is the same, and it reads as desperate to a human, which is the more immediate cost.

What a link cannot fix

Three honest limits. A link cannot make a post worth reading; if the destination answered a question nobody asked, routing more people to it only distributes the disappointment more efficiently. That is the failure we keep returning to — Google's problem was never machine authorship, it was content nobody needed — and link structure does not change that arithmetic.

An engine also cannot know what is about to change. It doesn't know that the pricing post goes stale next Tuesday, or that your sales team quietly abandoned the framing in a post from March. Those are facts about your business, and they arrive through a human or not at all.

And automated linking overdoes it. Left unsupervised, a system that can find a plausible connection finds one in every paragraph, and nine hundred words carrying eleven internal links reads like an encyclopedia stub. We cap it deliberately. Three good links beat eleven defensible ones.

The archive is an asset only once it is connected

Publish weekly for two years without linking and you own a hundred separate pages. Link as you go, and go back, and you own one thing with a hundred entrances.

That difference shows up wherever a reader lands deep instead of on your homepage, which is now most of them. Someone arriving from a search result or an AI-generated answer sees exactly one page of your site plus whatever that page offers them next. For that reader, your internal links are the entire navigation.

It costs nothing. It requires no new content. It is the highest-return hour in SEO, and it stays undone because it is the one job that asks you to remember everything you have ever published.