Internal Linking Strategy: The Expert Plan, From Priority Pages to Multilingual Sites

Quick quiz

Which way of finding customers online fits you best?

Question 1 of 425%

How do your customers find businesses like yours?

They search on Google or ask ChatGPTThey discover them on social mediaSomeone recommends them

This guide goes further than the definition. If you only need the basics of what an internal link is, read our definition of internal linking first. Here we cover the plan: how link value moves through a site, how to choose the pages that deserve it, the three main linking models with their trade offs, a step by step method, the numbers to track and the mistakes that quietly cost rankings, including the ones specific to multilingual sites.

What an internal linking strategy is (and what it is not)

Most sites add internal links one article at a time: a writer remembers an older post and links to it. That is a tactic. A strategy works the other way round. You decide first where you want authority to end up, then every new link serves that goal.

A real strategy answers four questions in writing:

1. Which pages matter most for the business, and need the most internal links?
2. Which pages belong together by topic, and how do they link to each other?
3. Which anchor text points to each important page, and which anchors are forbidden?
4. Which linking model the site follows, so that every editor and every template applies the same logic?

Why search engines care about internal links

Internal links do three jobs at once. They let crawlers discover pages. They pass a share of authority (the logic behind PageRank). And their anchors and surrounding text tell Google what the target page is about. Google's own documentation on crawlable links puts it plainly: "Every page you care about should have a link from at least one other page on your site." The same page adds that there is no ideal number of links per page.

AI search engines rely on the same signals. Their crawlers follow links to find pages, and a page buried six clicks deep is rarely fetched, so it is rarely cited. A clean internal structure helps both Google and the answer engines.

What the data says

The largest public study on the topic comes from Cyrus Shepard at Zyppy (updated February 23, 2026). He analysed 23 million internal links across 1800 websites and compared them with Search Console clicks. It is a correlation study, so read it as a direction, not a law. The direction is still clear.

Two other findings matter for your plan. Traffic stopped rising past roughly 45 to 50 links, most likely because those extra links were sitewide menu links pointing to weak pages. And the variety of anchors pointing to a page correlated strongly with more clicks. You can read the full Zyppy internal link study for the charts.

How link value really moves through a site

Before you plan, you need a mental model of how value flows. Simplified: each page has a certain amount of authority (from backlinks and from the internal links it receives). It shares that authority among the links it contains. A link placed in the main text, likely to be clicked, counts more than a link hidden in a footer. This is the "reasonable surfer" idea described in a Google patent: links are not all weighted equally.

Click depth

Click depth is the number of clicks from the home page to a page. The home page usually holds the most authority, so every extra level dilutes what reaches the page. A common rule is to keep important pages within three clicks. Depth also affects crawl frequency: on large sites with a tight crawl budget, deep pages are visited less often and updated later in the index.

The five types of internal links

Not every link has the same role. Your strategy should say what each type is for.

Link type	Where it sits	Main role	Weight as a signal	What to decide
Contextual link	Inside the body text	Pass relevance and authority between related pages	High: surrounded by topical text, likely to be clicked	Which pages link where, with which anchor
Main navigation	Header menu, on every page	Give top sections a link from every URL	Medium: sitewide, little context	Only 5 to 9 top sections, never long tail pages
Breadcrumb	Above the title	Push value up to parent categories, show the hierarchy	Medium: structured, consistent anchors	Path logic and the BreadcrumbList markup
Related content block	End of article or product page	Connect siblings in a cluster	Medium: can be automated	Rules: same cluster first, then neighbouring clusters
Footer	Bottom of every page	Legal, contact, a few key pages	Low: Google discounts boilerplate	Keep it short, avoid 80 keyword links

The practical consequence: when you want to boost a page, contextual links from strong, related pages are your first lever. Menu links are the second, and they are a scarce resource.

Pillar 1: choose your priority pages

A strategy starts with a short list. If every page is a priority, none is. You want 10 to 30 pages per site section that receive the bulk of your internal links: the pages that sell, the pages that already rank near the top, and the guides that support them.

Where to find them

Cross three sources:

• Business value: product, service, pricing and demo pages. These are the pages you want people to land on.
• Search Console: queries ranking in positions 4 to 20 with real impressions. A few strong internal links are often enough to move a page from position 8 to position 4. Filter the Performance report by page, sort by impressions and keep pages with an average position between 4 and 20.
• Backlinks: pages that receive many external links. They hold authority you can redistribute to the pages that need it. A page with 40 referring domains and no outgoing contextual links is wasted potential.

Give each priority page a role and a target

For each page on the list, write three things in your plan: the main query it targets, the cluster it belongs to, and a target number of contextual inlinks. A workable starting grid: conversion pages and pillar guides at 10 or more unique contextual inlinks, cluster articles at 3 to 8, utility pages at 0 to 2. These are working targets, not rules from Google. Adjust them to the size of your site: a 60 page site cannot give 15 links to 20 pages.

The pages that act as anchors of your site are often called cornerstone content. Our guide to cornerstone content explains how to choose and write them so they deserve the links you send.

Pillar 2: build topic clusters

The second pillar groups pages by topic. A cluster is a pillar page that covers a broad subject, surrounded by narrower articles that each answer one specific question. Pillar and articles link to each other. The result: Google sees a block of pages that cover a subject in depth, which builds topical authority, and each page passes relevance to its neighbours.

The linking rules inside a cluster

• Pillar to articles: one contextual link per article, in the section that introduces its subtopic.
• Articles to pillar: one link, early in the text, with an anchor close to the pillar's main query.
• Article to article: only when the reader genuinely needs the other page next. Two or three sibling links per article are plenty.
• Across clusters: allowed, but deliberate, from a sentence that bridges the two subjects.

Grouping queries by shared Google results is the core of keyword clustering, and the full pillar and spoke model is detailed in our guide to topic clusters and pillar pages. If you want a map produced for you, the Sorank topic cluster generator groups keywords into clusters and flags overlaps that would cause cannibalization.

Pillar 3: set your anchor text rules

The anchor is the clickable text of a link. Google reads it as a short description of the target page. Its documentation asks for anchors that are "descriptive, reasonably concise, and relevant" to both the page they sit on and the page they point to, and warns against generic text such as "click here" and against keyword stuffing.

The rules to write down

1. One main anchor per priority page, close to its target query, used in some of the links (not all).
2. Variations around it: partial matches, synonyms, questions, the page title. The Zyppy data links anchor variety with more clicks.
3. Never the same anchor for two different pages. If "internal linking audit" points to page A in one article and to page B in another, you tell Google both pages answer the same query. That is a direct path to keyword cannibalization.
4. No generic anchors on contextual links: "here", "this article", "read more".
5. Anchors read well on their own. Google suggests a simple test: read the anchor alone. If it does not tell you what the target is about, rewrite it.

In code, the difference looks like this. The last example matters for image links: Google uses the alt attribute as anchor text.

<!-- Weak: generic anchor, no context -->
<p>To learn more, <a href="/blog-geo-seo/site-architecture-seo">click here</a>.</p>

<!-- Strong: descriptive anchor inside a relevant sentence -->
<p>Pages buried too deep get crawled less often, so check
your <a href="/blog-geo-seo/site-architecture-seo">crawl depth by directory</a> first.</p>

<!-- Image link: the alt text becomes the anchor -->
<a href="/pricing"><img src="/img/plans.png" alt="Sorank pricing plans"></a>

For a full list of anchor types (exact, partial, branded, naked URL, generic), see our guide to anchor text.

The three internal linking strategies compared

Once your pillars are defined, you choose how pages connect. There are three main models. Most real sites combine them, but one should lead, otherwise every editor follows a different logic.

1. The "Wikipedia" strategy: link every related page

Principle. Every time a page mentions a concept that has its own page, it links to it. Links go in all directions between all close pages, like on Wikipedia, where an article can contain hundreds of internal links.

When to use it. Large knowledge bases, glossaries, documentation, media sites, any site where readers jump from concept to concept.

Advantages. Very low click depth, fast discovery of new pages, many anchor variations, excellent for readers who explore. It is simple to apply: if a term has a page, link it.

Limits. Authority spreads evenly, so you cannot push a few priority pages. On a small site with little authority, the dilution hurts. Topic boundaries blur, so Google gets a weaker signal about what each section covers.

Example. A glossary entry on "crawl budget" links to "crawling", "robots.txt", "sitemap", "log files" and "server errors" in the first paragraphs, and each of those pages links back.

2. The semantic cocoon: mother, daughter and sister pages

Principle. Popularised in France by Laurent Bourrelly under the name "cocon sémantique", the cocoon organises pages as a strict hierarchy built from search intent. A mother page covers a broad intent. Daughter pages cover narrower intents below it. Pages at the same level are sisters. Links follow precise rules: the mother links down to all its daughters; each daughter links up to its mother; a daughter links to its sisters only when the topic flows naturally; and nothing links outside the cocoon except through the top page.

When to use it. Competitive niches where topical authority makes the difference, sites you build from scratch, sections you can plan fully before writing.

Advantages. The strongest relevance signal of the three models. Authority is concentrated where you want it, because links climb back towards the mother page. Every page has a clear role, which prevents cannibalization.

Limits. Heavy to plan and maintain. Strict rules can force artificial links or forbid useful ones. Hard to retrofit on an existing site with years of unplanned links. Purists ban links outside the cocoon, which can hurt the reader.

Example. Mother page: "running shoes". Daughters: "trail running shoes", "road running shoes", "running shoes for flat feet". Granddaughters below "trail running shoes": "waterproof trail shoes", "trail shoes for mud". The waterproof page links up to the trail page and sideways to the mud page, never directly to the road page.

Our full guide to the semantic cocoon covers how to build one level by level.

3. The URL silo: structure by directory and breadcrumb

Principle. The hierarchy is written into the addresses. Each section has a directory, each page sits in the directory of its topic, and the breadcrumb mirrors that path. Links mostly stay inside their silo: pages link to their parent directory page and to siblings in the same directory.

/seo/                                  section page (pillar)
/seo/technical/                        category page
/seo/technical/crawl-depth/            article
/seo/technical/orphan-pages/           article
/seo/content/                          category page
/seo/content/topic-clusters/           article

When to use it. Online stores, catalogues, local service sites, any site with a natural tree of categories. It also suits teams with many editors, because the rule is visible in the address itself.

Advantages. Easy to understand for everyone. The breadcrumb creates consistent upward links automatically. Reports by directory (crawl, traffic, indexing) become trivial. It is the model most site architectures already follow, as explained in our guide to site architecture for SEO.

Limits. A page can only live in one directory, while real topics overlap. Moving a page later means changing its URL and adding a redirect. Strict silos block useful cross links between sections.

Add breadcrumb markup so Google understands the path. Google's breadcrumb structured data documentation requires at least two ListItems, each with a position and a name:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "BreadcrumbList",
  "itemListElement": [
    {"@type": "ListItem", "position": 1, "name": "SEO", "item": "https://www.example.com/seo/"},
    {"@type": "ListItem", "position": 2, "name": "Technical SEO", "item": "https://www.example.com/seo/technical/"},
    {"@type": "ListItem", "position": 3, "name": "Crawl depth"}
  ]
}
</script>

Comparison table

Criterion	Wikipedia mesh	Semantic cocoon	URL silo
Core rule	Link every related concept	Links follow the intent hierarchy	Links follow the directory tree
Best for	Glossaries, docs, media	Competitive niches, new sites	Stores, catalogues, local services
Authority flow	Spread evenly	Concentrated on mother pages	Concentrated on category pages
Topical signal	Weak to medium	Strong	Medium to strong
Effort to plan	Low	High	Medium
Effort to maintain	Medium (many links)	High	Low
Retrofit on an old site	Easy	Hard	Medium (URL changes)
Main risk	Dilution, no priorities	Artificial links, rigidity	Isolated sections

Our recommendation for most sites: use the URL silo or the cocoon as the backbone (it decides where authority ends up), and allow a light mesh inside each cluster (it keeps pages useful for readers). Keep cross silo links rare and intentional.

Step by step: build and roll out your plan

Here is the method we use, from audit to rollout. On a site of a few hundred pages, plan two to three working days for the first pass.

Step 1: crawl and export

Use a crawler such as Screaming Frog SEO Spider. The free version crawls up to 500 URLs, enough for many small business sites. Our Screaming Frog setup guide covers the configuration.

In the Internal tab, four columns drive your plan. Screaming Frog's documentation defines them: Inlinks (number of internal links to the URL), Unique Inlinks (a page linked three times from the same page counts once), Crawl Depth (clicks from the start page) and Link Score (a 0 to 100 metric of the relative value of a page based on its internal links, computed after a crawl analysis).

Step 2: find orphan pages with a formula

A crawler only finds pages that are linked. To find orphans, compare a list of all known URLs (XML sitemap plus Search Console Pages export) with the crawl. In Google Sheets, put all known URLs in column A of a "Known" sheet and the crawled URLs in column A of a "Crawl" sheet, then:

=FILTER(Known!A2:A, Known!A2:A<>"", ISNA(MATCH(Known!A2:A, Crawl!A:A, 0)))

Every URL returned is known to Google or to your sitemap but receives no internal link. Decide for each one: link it from its cluster, merge it into a stronger page, or remove it with a redirect.

Step 3 to 5: fix structure, then add links from strength

Sort your crawl by Link Score or by referring domains. Your top 20 pages are your donors. For each priority page, find two to five donors on a related topic and add one contextual link each, with an anchor from your plan. Placing the link in the first half of the text, inside a sentence that explains why the reader should click, follows the reasonable surfer logic.

Step 6: remove internal redirects

When a URL changes, old internal links keep pointing to the redirect. Each hop wastes crawl and can lose a little signal. After every migration, export internal links whose target returns a 301 or 302 and replace them with the final URL. Our guide to SEO redirects explains the status codes, and the free broken link checker finds dead links on any page.

How to measure an internal linking strategy

Track a small set of numbers, the same way every month, on your priority pages first.

Metric	Where to read it	What good looks like	Action if it is off
Unique contextual inlinks	Crawler, Internal tab	At or above the target in your plan	Add links from donor pages
Crawl depth	Crawler	3 or less for priority pages	Link from the menu, a hub or a category page
Orphan pages	Sitemap and Search Console compared with crawl	Zero among pages you want indexed	Link, merge or redirect
Internal links to redirects or errors	Crawler, response codes of link targets	Zero	Replace with the final URL
Average position and clicks	Search Console, Performance by page	Rising 4 to 12 weeks after the links went live	Check anchors and the content itself
Pages indexed	Search Console, Pages report	Priority pages all indexed	Inspect the URL, check links are crawlable

Note the date of every batch of new links in your plan. Without it, you cannot tell in Search Console whether a ranking change came from links, from content or from an algorithm update.

Common internal linking mistakes (and how to fix them)

Each mistake below is common on real sites. For each one: what it is, how to spot it, how to fix it.

1. Orphan pages

What it is. A page with no internal link pointing to it. Google may still know it through the sitemap, but it receives no authority and little crawl.

How to spot it. The sitemap versus crawl comparison from step 2.

How to fix it. Link it from its pillar and from one or two siblings. If no page naturally links to it, it probably should not exist alone: merge it.

2. Priority pages buried too deep

What it is. A money page five or six clicks from the home page, often because it only appears on page 4 of a paginated blog.

How to spot it. Filter the crawl by depth greater than 3 and cross with your priority list.

How to fix it. Add it to a hub or category page, to the relevant menu, or link it from a strong article. Avoid relying on pagination alone to reach content.

3. The same anchor pointing to several pages

What it is. Two different URLs receive the same anchor, so Google cannot tell which one answers the query.

How to spot it. Export all internal links (anchor and destination), pivot by anchor, count distinct destinations. Any anchor with more than one destination is a flag.

How to fix it. Reassign each anchor to one page only. If both pages really answer the same query, merge them.

4. Links Google cannot follow

What it is. Navigation built with buttons, onclick events or framework components that never render a real link. Google states that it reliably follows <a> elements with an href attribute, and that markup such as <span href>, <a onclick> or javascript: URLs is not reliable.

How to spot it. Crawl with JavaScript rendering off, then on. Pages found only with rendering on, or not at all, depend on fragile links. The URL Inspection tool in Search Console shows the rendered HTML.

How to fix it. Render every navigational link as <a href="/blog-geo-seo/site-architecture-seo">. Our guide to JavaScript SEO covers the rendering side.

5. Nofollow on internal links

What it is. Adding rel="nofollow" to internal links to "sculpt" authority. It does not keep the authority on the page, it simply stops it from flowing through that link.

How to spot it. In your crawler, filter internal links by the nofollow attribute.

How to fix it. Remove nofollow from internal links. If a page should not get authority, remove the link or noindex the page. See our page on nofollow links.

6. Bloated menus and footers

What it is. A mega menu with 150 links or a footer listing 80 keyword pages. Every page then links to everything, which flattens your priorities. The Zyppy data also shows traffic falling past roughly 50 links, likely from this kind of sitewide link.

How to spot it. Count links in the header and footer templates. If the sitewide links outnumber the contextual ones on most articles, the template dominates.

How to fix it. Keep 5 to 9 top level entries in the menu, move long lists to hub pages, and keep the footer for legal, contact and a handful of key pages.

7. Internal links between languages that do not match

Multilingual sites add a family of mistakes that most guides skip. Language versions are separate pages in Google's eyes, and links tell Google which versions belong together. Our guide to international SEO covers the wider setup; here are the linking errors.

7a. A page in one language linking to a page in another. A French article links to the English version of a guide, often because a translator kept the original link. The French reader lands on English content and leaves, and the French guide loses an internal link it should have received.

How to spot it. In Screaming Frog, run Bulk Export, Links, All Inlinks, then filter rows where the source is in one language folder and the destination in another. With folders such as /fr/ and /en/, these two regular expressions isolate French pages linking to any other language folder:

Source:      ^https://www\.example\.com/fr/
Destination: ^https://www\.example\.com/(?!fr/)[a-z]{2}/

How to fix it. Replace each link with the equivalent page in the same language. The only cross language links that belong in body content are the language switcher and deliberate references ("this study is only available in English").

7b. Versions that do not point to their translation. The French page has an English equivalent, but nothing connects them: no hreflang, no language switcher pointing to the exact translation (the switcher goes to the English home page instead). Google may then treat them as unrelated pages, or show the wrong language in results.

How to spot it. Check the language switcher on a sample of deep pages: does it go to the translated page or to a home page? Then check hreflang tags on the same pages.

How to fix it. Make the switcher link to the exact equivalent URL, and declare every version with hreflang. Google's documentation on localized versions is strict: "Each language version must list itself as well as all other language versions", and "if two pages don't both point to each other, the tags will be ignored."

<!-- On BOTH the English and the French page -->
<link rel="alternate" hreflang="en" href="https://sorank.com/blog-geo-seo/site-architecture-seo" />
<link rel="alternate" hreflang="fr" href="https://sorank.com/fr/blog-geo-seo/architecture-site-seo" />
<link rel="alternate" hreflang="x-default" href="https://sorank.com/blog-geo-seo/site-architecture-seo" />

7c. Hreflang and internal links that tell different stories. The hreflang says page A in French matches page B in English, but the canonical of page B points elsewhere, or the switcher links to page C, or the French internal links point to a redirected URL while hreflang lists the final one. Google receives contradictory signals and may ignore the hreflang cluster.

How to spot it. For a sample of pages, write four URLs side by side: the hreflang target, the switcher target, the canonical of the target page, and the final URL after redirects. All four must be identical.

How to fix it. Generate hreflang, canonical and switcher links from the same source of truth in your CMS (one translation table), never by hand. The free Sorank hreflang generator builds a correct, reciprocal set of tags, and our hreflang guide explains each attribute.

8. Linking once and never again

What it is. New articles get links to older ones, but old articles never link to the new ones. Fresh pages stay weak for months.

How to spot it. Sort pages by publication date and look at unique inlinks for the last three months of content.

How to fix it. Make it part of publishing: every new article gets two to five links from existing, related pages on the day it goes live.

Scaling internal linking without losing control

Manual linking works well up to a few dozen pages. Past a few hundred, nobody remembers every page, and links start pointing to outdated or cannibalizing content. Automation helps, with three safeguards:

• The plan stays human. Tools can suggest links, but the priority list, the clusters and the anchor rules are business decisions.
• Relevance before volume. An automated link must come from a sentence that actually talks about the target topic. Reject suggestions that only match a keyword.
• Review in batches you can undo. Apply links per cluster, note the date, check the impact, then continue.

This is where Sorank fits. It plans topic clusters with anti-cannibalization checks through its topic cluster generator, and its content optimization feature rewrites and updates older pages, which is the moment to refresh their internal links.

What to do today

You do not need to rebuild your whole site to start. In one afternoon:

1. Crawl the site and export inlinks, unique inlinks and crawl depth.
2. List 10 priority pages from Search Console (positions 4 to 20) and from your offer.
3. Run the orphan formula and link or merge every orphan you want indexed.
4. For each priority page, add three contextual links from your strongest related pages, with anchors from a written list.
5. If your site is multilingual, run the cross language regex and fix every mismatched link.
6. Note today's date in your plan and recheck positions in six weeks.

Then decide which model leads your site (silo, cocoon or mesh) and write it down. Once the model and the anchor rules are written, every new page fits into the structure instead of drifting outside it.