Internal linking: the lever almost nobody uses well

In a 312-page recipe site I audited, the two pages with the most impressions received not one link from inside the site itself. One had 77,000 impressions a quarter and eight clicks. The other, 29,000 impressions and 221 clicks. Zero inbound internal links, both. There were 35 pages like that, roughly 180,000 quarterly impressions the site's own structure simply never reached. Nobody had broken them: they had never been connected. The variable that decides whether your internal linking works is not how many links you have, it is where they come from.

How many internal links should a page have?

Six to ten in a standard article, and ten to sixteen in a pillar. On a new site it matters little because there is little strength to pass around; on a mid-authority site it is the lever that moves the most; on a consolidated site the work becomes maintenance and pruning links that no longer describe anything.

Quick answer

• The first thing to look at is not the links, it is the orphans. A page with no inbound links depends on Google remembering it.
• Footer and menu links barely count. In that site, 79 % of the existing links were there, and most pages had none inside the text.
• The rule that orders the work: do not write a sentence to place a link. Find mentions that already exist and wrap them.
• If removing the link leaves the sentence unsupported, it was well placed. If nothing changes, it was filler.

What kind of problem this is

The usual mental model is that internal linking "passes authority", and that phrase, from so much repetition, stopped meaning anything. It leads to placing links by quantity: a related block, a list of highlights in the footer, four links at the end of every article.

An internal link does three different things and only the first is automatic. It tells Google the page exists and how to reach it. It tells Google what it is about, through the anchor text and what surrounds it. And it tells Google how much importance you give it within your own site. The last two depend on where the link sits. A footer link repeated on 312 pages does not say the page is important: it says you have a footer. One inside a paragraph that explains something says much more, because the context backs it.

How much work lies ahead depends on the state of the site. A new site under twenty pages has no linking problem yet; it has nowhere to link to, and the only thing that matters is not creating orphans from the start. A mid-size site with history, 50 to 500 pages, is where the money is left on the floor: there are almost always orphans, the bulk of the links sit in repeated templates, and the highest-impression pages are the worst connected because they were published early and nobody went back. A large site over 500 pages has a different problem: not missing links but missing hierarchy, everything linking to everything and no page standing out. There the work is to prune, not to add.

How it gets fixed, in order

1. Find the orphans before anything else

An orphan is a page with no inbound link from the site itself. You find it by crossing two lists: all your URLs (from the sitemap) and all the URLs that receive an internal link (from a crawl with Screaming Frog's free version, which reaches 500 pages). Order matters: sort the orphans by impressions, not alphabetically. In the case that opens this article, the top two on that list held more than 100,000 impressions between them. That a page has no internal links does not mean it cannot be crawled, a myth taken apart in why a page is not indexed; it means it receives none of the strength the rest of the site distributes.

2. Look at where the links you already have sit

If 79 % of your internal links live in the footer and the menu, you do not have a quantity problem. The check is fast: open three articles and count the links inside the paragraphs. If the answer is zero, you know where to start.

The three series sit in separate panels because they measure different things, and putting them on one axis would flatten the orphans and the broken links against the floor beneath the 367 links. The middle one is what changed the business: 35 pages holding some 180,000 impressions a quarter that the internal structure never reached.

3. Find mentions that already exist

This is the rule that orders the work, and it runs against instinct: do not write a sentence to place a link. Instead, search the existing content for the times you already mention the topic of the page you want to link, and wrap that mention. In the project cited, 257 such mentions were found. 116 were applied and 141 discarded: 109 because the anchor was false (the word appeared in another sense) and 32 because they fell inside a FAQ block, where a link breaks the answer an engine will extract. Discarding 55 % is not a failure. It is what separates linking that reads natural from linking that reads as filler, and why the extractable answer matters is in how to get cited by ChatGPT.

4. Write anchors that say something

The anchor is the label Google reads the destination by. "Click here" describes nothing; "structured data guide" does. The opposite extreme also hurts: repeating the exact same anchor across twenty inbound links reads as generated, because it is. Two or three variants per destination page is enough. How the same discipline applies to the page's own writing is in how to write SEO content.

5. Close the loop with every publication

The operating rule that prevents orphans returning: every new article receives at least one inbound link from an older one within 24 hours. You edit the old article, not the new one. It costs five minutes and it is the difference between a site that links itself and one that has to be rescued every two years. Where new articles that link to unpublished ones go, and why that is not a 404, is a separate mechanic; earning external links to the pages you connect is in how to get backlinks.

When NOT to touch internal linking

When the problem is something else. In that same recipe site there were pages at position 1 with zero clicks, and no internal link was going to fix that: the title Google showed did not contain the term people typed. Whether your problem is positions or clicks is diagnosed differently, in why organic traffic drops, and whether the query is even yours is read off search intent.

Two things repeated that are not true

"An orphan page cannot be crawled or indexed." False, and widely repeated. A page with no internal links is crawled perfectly well if it is in the sitemap, if someone links it from outside, or if indexing is requested by hand. What happens to an orphan is not that it is invisible: it is that it receives none of the strength the rest of the site distributes and its crawl priority drops. Two real problems, and neither is the one stated.

"I added eight internal links and the article reached the first page in six weeks." The kind of case that circulates as a technique. It may have happened exactly like that and it is still not a technique: it is one case, uncontrolled, on a specific site, at a moment when other things were also happening. This blog does not promise ranking timelines, because nobody can isolate the cause in a sample of one. What can be said, from sweeping whole sites, is duller: the pages that receive zero internal links are usually the ones the business needs to work, because templates link the generic and nobody links the specific by hand.

Mistakes that repeat

• Counting links without looking at where they come from. Twelve in the footer are worth less than two inside the text.
• Writing sentences to place links. It shows, and turns the article into a directory.
• Always linking to the same place. If every article points at the services page, no other page exists for Google.
• Forgetting the inbound link at publication. That is how orphans are created, one at a time, unnoticed.
• Using the exact same anchor every time. An identical anchor profile reads as automated.
• Fixing one broken link and moving on. If there is one, there are nine.

Data and transparency

That an orphan page cannot be crawled or indexed is a false claim appearing in several of the first results, checked on 12 August 2026: a page with no internal links is crawled if it is in the sitemap, linked from outside, or submitted by hand. What it loses is internal strength and crawl priority. No maximum number of links per page appears here: the ones that circulate (45, 50, 100) have no published source. No timed improvement case is recorded, because a case without a control is not a technique. The figures come from an own audit of a Colombian recipe site in English, 312 pages on WordPress, run in 2026. The site is client work and is not named; figures are rounded. Documented result: in-prose internal links from 251 to 367, orphan pages from 35 to 0, broken internal links from 9 to 0. Verified as of September 2026.

What this changes

When someone asks about internal linking they are almost always asking how many links to place, and that is the part that matters least. The useful question is another: if Google forgot your site tomorrow and could only travel it by following links from the home page, how many of your pages would it reach?

In this article's site, the answer was 277 of 312. The 35 missing produced 180,000 impressions a quarter and were there by luck. Nothing new had to be written and not a single external link earned: what was needed was to connect what already existed.

Frequently asked questions