The 1,000-page site that barely links to itself

We measured a business site we help maintain. About 1,200 URLs in the sitemap — 986 of them blog posts, a decade of newsletters and updates, published diligently, month after month.

Then we sampled fourteen of those posts and counted the links a writer had actually put inside a sentence. The average was 1.1 per post. Six of the fourteen had none at all.

Every one of those pages looked healthily linked at a glance, because the template adds seven or eight links to every page automatically — navigation, categories, related posts. That's the trap. The count looks fine; the site is a field of orphans.

The illusion, measured

Two counts, same fourteen pages. The faded bars are links the template puts on every page whether or not anyone thought about it. The solid bars are links somebody chose to place inside a sentence, pointing at a specific related page, because it was genuinely useful.

The gap between those two numbers is the gap between a site that has links and a site that is linked.

This matters because the two kinds of link do different work. Template links are structural — they help crawling, and every page has the same ones, which is precisely why they don't distinguish anything. A link that appears on all 1,200 pages tells a search engine nothing about which pages matter. Contextual links are editorial judgement made machine-readable: this page is related to that page, specifically, and here's the phrase that connects them.

Why a decade of publishing can produce almost nothing

The pattern behind this is common and completely understandable.

A business commits to publishing. Someone writes a monthly newsletter, an industry update, a staff spotlight. It goes up. The next month, the same. Ten years later there are a thousand posts, each one written as a standalone artefact, none of them aware the others exist.

Nothing was done wrong at any single step. But the result is a library where every book is on its own shelf in its own room, and the only route to any of them is the front desk.

Three consequences, in rough order of expense:

Authority doesn't circulate. When another site links to one of your pages, that value can flow to related pages through internal links. With no contextual links, it stops where it lands.

Nothing signals importance. Search engines infer what matters partly from what you link to repeatedly with meaningful anchor text. If your key service page is linked only from the navigation, you're saying it's one of forty equally important things.

Readers leave. The person who arrived on a 2021 newsletter about workplace safety is interested in workplace safety. If nothing on that page offers them the next thing, they read and go — and you paid to produce that page.

Three shapes a large site can have

Most content-heavy business sites are the first shape and would benefit enormously from becoming the third. The middle shape — hub and spoke — is the achievable intermediate step, and usually the right target for a first pass.

Orphan field. Everything hangs off the navigation. Nothing links sideways. This is what publishing without a linking habit produces.

Hub and spoke. A hub page for each topic links out to everything relevant, and each page links back to its hub. This alone fixes most of the problem, because it creates a route between related pages that isn't the global nav.

Connected cluster. Hubs, plus lateral links between related pages. This is the target, but it's earned incrementally rather than built in one pass.

We've written about how topic clusters work structurally; this article is about retrofitting one onto a site that already has a thousand pages and no structure.

How to audit your own site in an afternoon

You don't need enterprise tooling. You need a crawler and a spreadsheet.

Step three is the one people skip, and it's the one that makes the audit truthful. If you count every link on the page, a well-templated site looks beautifully connected — 8.6 links per page in our sample — and you'll conclude there's no problem. Excluding links that appear site-wide is what reveals the real number.

The quickest version of this check: open three of your own blog posts and count the links inside the paragraphs. If it's zero or one, you have this problem, and you almost certainly have it a thousand times over.

What to fix, in order

Retrofitting links across a thousand pages by hand is not a project anybody finishes. Do it in this order and stop when the returns flatten.

1. Build the hubs first. One page per major topic or service, which links to everything relevant. If you have five service lines, that's five pages. These are the pages that will rank for the valuable terms, and they give every future link somewhere sensible to point.

2. Link the top twenty pages by traffic to the relevant hub. These already earn attention. Passing some of it to a commercial page is the highest-value hour of work available.

3. Link the pages with backlinks. Anything another site references should route that authority onward. Frequently a handful of pages, frequently the most valuable edit in the whole programme.

4. Fix the money pages' inbound count. For each service or product page, deliberately find ten older posts that could reasonably link to it, and add a sentence to each that does. Not a widget — a sentence, with real anchor text.

5. Add a linking step to publishing. From now on, no post is finished until it links to at least two existing pages and one existing page links to it. This is the single change that stops the problem regenerating.

6. Only then consider the long tail. The 800 remaining posts can be handled in batches, or left alone. Many of them are newsletters from 2019 and honestly aren't worth the effort — which is itself a finding worth acting on.

Step five is the important one. Everything else is remediation of a habit; step five is the habit changing.

The methodology, so you can repeat it

The measurement above is simple enough to reproduce on any site, and worth describing precisely because the definition of "internal link" is what determines whether the answer means anything.

We pulled the sitemap index and counted URLs per sub-sitemap — posts, pages, staff profiles, categories, tags. That gives the true size of the site, which is almost always larger than the organisation believes. A business that thinks of itself as having "about forty pages" often has a thousand once a decade of posts is counted.

Then we sampled fourteen posts at random and, for each, extracted the main content container and counted two things separately: every internal link inside that container, and only those internal links appearing inside a paragraph, excluding category, tag, author and file links.

That second definition is the useful one. A link inside a sentence was placed by a person who decided two pages were related. That is the editorial signal — and it is the thing that a template cannot manufacture.

Both numbers matter, but only when reported separately. Together they average out into a figure that reassures everyone and describes nothing.

A caveat on our own measurement: fourteen pages is a sample, not a census, and content-container detection is imperfect across templates. We'd treat the numbers as indicative of the pattern rather than precise. The pattern, though, was unambiguous — no sampled page had more than four contextual links, and nearly half had none.

Why this is worth more than it sounds

Internal linking has an image problem. It sounds like housekeeping, it produces no artefact anyone can admire, and it is impossible to make exciting in a proposal. So it gets deferred in favour of a redesign, which is visible.

Three reasons it outperforms most of what it gets deferred for:

It uses pages you have already paid for. A thousand posts represent a decade of salary. Linking them is the cheapest possible way to get more from work that is already done and currently earning nothing.

It compounds. Each new link improves the pages at both ends, and improvements accumulate rather than depreciate. A design refresh looks its best on launch day; a linking structure looks its best in year three.

It's one of the few things you fully control. You cannot control competitors, algorithms or what people search. You can decide, entirely on your own, which of your pages point at which other pages.

There's a newer reason too. When an assistant answers a question by reading a site, a well-linked cluster gives it more to work with than an isolated page — the related material is reachable and evidently connected. It's the same structural work that has always helped readers and crawlers, and it now helps a third audience.

None of that requires a rebuild, a new platform, or a budget cycle. It requires a crawl, a list, and someone willing to spend an afternoon adding sentences.

A worked example

Say a business has forty posts about workplace safety, published over six years, and one service page selling safety consulting. Today the forty posts and the service page are strangers.

Build the hub. Create one page — "Workplace safety" — that is genuinely useful in its own right: what the risks are, what the obligations are, what a programme involves. Then have it link out to the best fifteen of those forty posts, grouped under sub-headings, with anchor text describing what each covers.

Link back. Add one sentence to each of those fifteen posts pointing at the hub. Not a banner, not a widget — a sentence in the body, at the point where a reader would want more.

Route to the money page. The hub links to the safety consulting service page, and so do the five posts most likely to attract someone with a budget.

Prune quietly. Of the forty, perhaps twelve are timeless, fifteen are dated but salvageable, and thirteen are newsletters from 2019 that nobody will ever read. Leave those thirteen alone, or consolidate the best parts into one page and redirect the rest.

That's roughly a day of work. What it produces: a page that can rank for the topic rather than the brand, fifteen previously isolated pages that now route attention somewhere commercial, and a clear structure for everything published afterwards.

Multiply by however many topics the business actually sells, and that's the whole programme. Five topics, five days, spread over a quarter. It is not glamorous and it reliably outperforms things that cost ten times more.

The objections, answered

"Isn't this just SEO busywork?" It's the opposite of busywork — it's the part of SEO with the clearest mechanism. You are deciding which of your pages are important and making that decision legible. If it feels like busywork, that's usually because it's being done as a bulk task rather than as editorial judgement.

"Won't too many links dilute the value?" In theory value divides across links on a page; in practice, at two to five contextual links, this is not the constraint anyone should be optimising. The sites with a real problem have too few, by an order of magnitude.

"Our CMS has an automatic related-posts plugin." Covered above — that's structure, not signal. Keep it for readers, don't count it as linking.

"We don't have time to edit a thousand posts." You shouldn't. Do the twenty that get traffic and the handful with backlinks. That's most of the available value and it's a day or two of work.

"Shouldn't we just rebuild?" Almost never for this reason — linking is a repair, and one of the cheapest available.

"Can't AI do this automatically?" Partly, and it's genuinely useful for the first pass — identifying candidate pairs and drafting the sentence. But the judgement of whether two pages are actually related, and whether a reader at that point in the paragraph would want that link, is the part worth a person. Bulk-generated links look exactly like what they are.

Anchor text, briefly

Since every link needs words:

• Describe the destination. "Our workplace safety programme" beats "click here" and beats "read more", both of which describe the action rather than the thing.
• Vary it naturally. The same exact phrase on 200 links looks manufactured. Write the sentence, then link the part of it that names the target.
• Link where the reader would want it, mid-paragraph at the point of relevance, not in a block at the bottom nobody reads.
• Don't link the same target twice in one page. The first one is the one that counts.
• Two to five contextual links per page is a healthy target. Thirty is a link farm.

What "related posts" widgets don't do

A reasonable objection: our site has a related-posts widget, so pages are linked automatically.

Automated related-post modules are template links. They usually appear site-wide, they're often generated from category or tag matching rather than genuine relevance, and they frequently sit below the fold beneath a share bar. They're worth having — they help readers — but they don't do the work of a contextual link, and counting them is how a site with a real linking problem persuades itself it doesn't have one.

The test is simple: would a human editor have chosen that link, in that place, for that reader? If a machine picked it because two pages share a tag, it's structure, not signal.

If you have a big site and no idea what shape it's in, send us the URL — the crawl takes an afternoon, and the orphan list is usually the most surprising document a marketing team sees all year.