Buzzmatic
SEO

Internal linking: the guide from basics to steering link equity

Portrait of Raoul Schreck, CEO & Co-Founder of Buzzmatic
AuthorRaoul Schreck
Published
Reading time40 min read
Back to all articles

Internal links decide which of your pages Google treats as important. Google's own documentation puts it this way: "Every page you care about should have a link from at least one other page on your site." (Google statement). Mueller called internal links "super critical for SEO" in 2022. Most websites still leave this lever untouched. This guide covers what we use in our projects: basics, strategy, deliberate link-equity steering and a method for measuring the results. Every claim is tagged with its evidence strength: evidence (test or study), Google statement, patent, leak or experience.

At a glance: Internal links work through three levers: link equity (how much authority a page passes on), topical relevance (how closely two pages are related) and anchor text (what the clickable text tells Google about the target page). Steering link equity by business value gives you an advantage that backlinks alone cannot deliver.
Explainer video, 5:21 min, 10 chapters: from “What is an internal link?” to “Your audit in three steps”.

Contents


Part 1: Basics

An internal link is a hyperlink that points from one page to another on the same domain. When buzzmatic.com links to buzzmatic.com/knowledge/, that is an internal link. When buzzmatic.com links to google.com, that is an external link. And when google.com links to buzzmatic.com, that is a backlink. If you just want the short version, see our knowledge article on internal linking. More on external links in our article on external links.

In HTML, an internal link looks like this:

1<a href="/category/running-shoes-women/">Women's running shoes</a>

Three parts are required: the <a> element, the href attribute with the target URL and clickable text (the anchor text). Optional attributes include title (hover text, not a ranking factor), rel (e.g. nofollow) and target (e.g. _blank for a new tab). Google only follows links with <a> and href (Google statement). Everything else (JavaScript navigation without href, onclick, <span> elements) may not be recognized.

A typical website has two groups of internal links:

Sitewide elements appear on every page: the main navigation (header), the footer and breadcrumbs. They carry a lot of link equity because they are present on every page.

Individual elements appear only on specific pages or page types: in-text links in body copy, in-content modules (teaser cards, product grids), related products (cross-links), filters and facets, sidebar links and contextual buttons (e.g. chips like "lightweight jackets" on a parent category).

For later analysis it helps to give each link element a shorthand (N = navigation, BR = breadcrumb, T = text link, C = content module, FL = filter, etc.) and document which template contains which elements. The shorthand reappears in the link matrix.

What is PageRank?

PageRank is the algorithm that set Google apart from other search engines in 1998. The idea comes from citation analysis: a scientific paper that is cited frequently is probably important. Brin and Page applied this principle to web links. Every link counts as a vote for the target page, and a vote from a page that has many or strong links of its own carries more weight (paper "The Anatomy of a Large-Scale Hypertextual Web Search Engine", 1998, primary source, historical).

The random surfer and the damping factor. The paper describes PageRank through a thought experiment called the random surfer: a user starts on a random page and clicks links. With a probability of 85%, they follow one of the links on the current page. With 15% probability they get bored and jump to a random page instead. The first value (d = 0.85) is called the damping factor (primary source, historical).

A quick calculation. Take a homepage with a value of 100 and three outgoing links. Each link carries 0.85 x 100 / 3 = roughly 28 points. The remaining 15 (i.e. 1 minus d, the jump probability) do not disappear. In the original formula they are distributed evenly across all pages as a base value that every page receives (primary source, second paper by Page, Brin, Motwani, Winograd, historical). The values are recalculated iteratively until they stabilize. The total stays the same. (Simplified view of the original 1998 PageRank. Google today uses weighted links and a seed-based variant; the exact formula is not public.)

Patent and license. The corresponding patent (US 6,285,999, inventor Lawrence Page) was owned by Stanford University and licensed to Google (SEC exhibit from Google's 2004 prospectus, primary source, historical). According to Google Patents it expired in 2018. Google still links to both the paper and the patent in its current ranking systems guide (Google statement).

Toolbar PageRank: 2000 to 2016. In December 2000 Google released a browser toolbar with a PageRank display: every page visited got a score from 0 to 10 (press release, historical). Already in 2011, Google said the public score differed from the one used internally and that PageRank was not a useful success metric (Google statement, historical). The last public update was in December 2013. From March 2016 Google removed the display entirely. Google never published a specific reason for the removal. The likely explanation is that the visible score had fuelled an obsession with PageRank and a market for buying links (interpretation).

PageRank still runs internally. Google's ranking systems guide states that PageRank has evolved significantly and remains part of the core ranking systems: "it continues to be part of our core ranking systems" (Google statement). The 2024 Google API leak confirms this: the production PageRank field is called PagerankNS (NearestSeeds), and the old Pagerank field is marked as deprecated (leak, weighting unknown). A conversation note from the US antitrust case against Google describes PageRank as "a single signal relating to distance from a known good source" (DOJ trial exhibit PXR0356, February 2025, primary source). Today's variant measures how close a page is, via links, to selected trusted seed pages (inference from leak + patent US 9,165,040, not confirmed by Google).

What does this mean for internal links? The core of PageRank has stayed: links distribute authority, and the homepage is the strongest internal source. The number of outgoing links on a page determines how much equity each one passes. Google distinguishes internal from external links (leak: the isLocal field marks internal links as a separate class). The sections in this guide build on these foundations: from steering link equity through the leak attributes to calculating internal PageRank.

Internal links serve four purposes at once:

1. Discovery and crawling. Google finds new pages primarily through links. Its own documentation says: "the vast majority of the new pages Google finds every day are through links" (Google statement). An XML sitemap supplements this but does not replace it. More on this in our article on crawling and indexing.

2. Link equity (PageRank). Google still uses a model based on PageRank, see What is PageRank? for the full background. Every page carries a certain weight and passes part of it through its outgoing links to the linked pages. The more internal links a page receives, the more important it appears to Google. The strength of this signal is divided by the number of outgoing links. A link from a page with few outgoing links can therefore pass more weight than a link from a page with hundreds of targets. The 2024 leak shows a dedicated attribute for this: homepagePagerankNs stores the homepage's PageRank value on every page of the site (leak, weighting unknown). More on the leak attributes in What the 2024 Google leak reveals about internal links.

3. Context through anchor text. The clickable text of a link tells Google what the target page is about. The Google documentation puts it this way: "Good anchor text is descriptive, reasonably concise, and relevant to the page that it's on." (Google statement) The SEO Starter Guide adds: link text "tells users and Google something about the page you're linking to" (Google statement). "Click here" as anchor text wastes this signal.

4. User navigation. Internal links guide visitors to related content and offers. Good internal linking increases pageviews per session and brings users closer to conversion.

In our work, we map every internal linking measure to one of three pillars:

Link equity controls how much authority a page receives. The lever: which pages link to which, how many outgoing links a page has, and whether the link appears on every page (sitewide) or only on one.

Topical relevance builds up when you link closely related pages to each other. A page about running shoes that links to brands, shoe types and buying guides builds topical depth. A link to the careers section does not.

Anchor text gives Google a keyword signal about the target page. "Salomon trail running shoes" is a good anchor text. "Click here" wastes the signal, and "running shoes trail running buy cheap online order" is keyword stuffing.

These three pillars help with prioritization: when you increase a page's link equity (more inbound links), strengthen its topical embedding (links from related pages) and sharpen its anchor texts, you have used the three strongest levers.

Your first five steps

If you have not been actively managing internal links, start here:

  1. Find orphan pages. Crawl your website with a tool like Screaming Frog and compare the found URLs with your XML sitemap. Pages that are in the sitemap but have no incoming internal link are hard for Google to reach.
  2. Check anchor texts. Export the anchor texts of your internal links (Screaming Frog: Bulk Export > Links > All Inlinks). Look for generic texts like "click here", "learn more" or empty anchors. Replace them with descriptive text that names the target page's topic.
  3. Check click depth of your most important pages. Which pages are most important for your business (highest revenue, highest search volume)? How many clicks does a user need from the homepage to get there? If it is more than three or four, create shortcuts: link boxes, teasers, cross-references in body copy.
  4. Retroactively link new content. When you publish a new blog post or category, link to it from existing, topically related pages. This step is almost always forgotten.
  5. Fix broken internal links. Internal links that point to 404 pages or redirects waste crawling resources and link equity. Search Console and every crawler report them.

Part 2: Strategy

Site architecture and click depth

The number of clicks a user needs from the homepage to reach a page is called click depth. Gary Illyes said in the Search Off the Record podcast in 2024: "from Google's perspective, the homepage is the most important page on the site" (Google statement). Pages closer to the homepage carry more weight (experience). URL structure (/shop/category/product/) matters less. See also our article on URL structure.

We keep revenue pages within three to four clicks (experience). The rule of thumb "no more than three clicks" is widespread, but there are few rigorous public tests to support it.

Hub and spoke, topic clusters and silos: Three architecture terms that mean roughly the same thing: group related pages and link them to each other. The pillar-cluster model comes from HubSpot (2017): a pillar article as an overview, linked to specialized cluster articles that link back. HubSpot popularized the model in 2017 but never published specific growth numbers (self-reported, no independent replication). Silos go a step further and recommend limiting cross-references between topic groups to the hub level. None of these models is backed by controlled tests. The underlying idea, linking topically related pages and concentrating link equity within a topic group, does align with what Google says about topical relevance. Our article on site architecture goes deeper on planning page structure.

Choosing anchor text

Google uses the anchor text of a link to understand the target page. This applies to external backlinks and internal links alike. The Zyppy study (updated February 2026) matched 23 million internal links across 1,800 websites (roughly 520,000 URLs) against GSC click data and found correlations with traffic (evidence, correlation):

URLs with at least one exact-match anchor text received at least five times more traffic than those without. More anchor text variants correlated strongly with more clicks. Empty anchor texts (over 6% of all links in the study) showed no statistical difference from the average. Naked-URL anchors (under 1% of links) correlated with almost 50% more traffic.

The authors emphasize that these are correlations. They call the results "directionally useful" and do not claim causation. A Zyppy follow-up study (updated September 2026, 50 sites) shows the flipside: websites with particularly high anchor text diversity per page were more likely to lose traffic after Google updates (correlation −0.337, evidence, small sample).

Our recommendation: Use descriptive anchor texts that name the target page's topic. Vary slightly ("women's running shoes", "running shoes for women"), but do not spread variations as widely as possible. Avoid "click here" and empty anchors.

Does the first link count? The state of "First Link Priority". Action rule: put your most descriptive anchor text in the first link to a URL in the HTML source. Two spots matter in practice. The navigation sits before the body content in HTML. If it links to the target page with a generic anchor like "running shoes", Google probably records that anchor first. On teaser cards, the image link usually comes before the text link in HTML and becomes the first link. Build the entire card as a single link with descriptive anchor text.

The idea has been circulating since tests in 2007/2008. Google has never confirmed it, but six documented tests support it:

  • 2008, Rand Fishkin (Moz): Three test domains, repeated experiment. With two links to the same URL, Google only counted the anchor text of the first (evidence, small test).
  • 2008, SISTRIX: One link pair with made-up anchor texts. Only the first anchor appeared in Google's cache of the target page (evidence, small test).
  • 2008, SEO Scientist: Swapped anchor order between tests. Result according to Search Engine Roundtable: Google valued whichever anchor came first (evidence, small test, secondary source).
  • 2008, SISTRIX Revisited: Five variants with image and text links. When the first link was an image link without usable alt text, Google counted the second text link's anchor. Nofollow on the first link caused Google to ignore both anchors (evidence, small test). Note: in this 2008 test, Google ignored the alt attribute of image links. Later tests and Google's current documentation contradict this (see Zyppy 2023).
  • 2011, YouMoz community: Test with five links to the same URL, including hash, parameter and 301 links. The target page ranked for the first link's anchor and additionally for the hash, parameter and 301 link anchors, which Google apparently treated as separate URLs (evidence, small test).
  • 2023, Cyrus Shepard (Zyppy): Analysis of the GSC report "Top Linking Text". Google reported one text anchor and one image anchor per target URL in most cases, often the first. Image alt text was captured here, unlike in the 2008 test (evidence, small test, sampled data).

Honest assessment: all tests used made-up words or sampled GSC data, with small sample sizes. Between 2012 and 2022 no documented test directly examined the rule. Whether anchor order measurably moves rankings for real keywords is unknown. The only documented counter-evidence from 2008 was considered methodologically weak in the discussion at the time.

Google's position: Mueller said in 2018 during Office Hours: "This isn't something we have defined, where we say 'It's always like this, it's always the first link, always the last link, always an average of the links.'" (Google statement). In 2026 on Reddit he advised against obfuscation: "I suspect you're overthinking it" (Google statement). Mueller said in 2025 that four identical links to the same page were "fine & common" (Google statement).

Same anchor, different targets. When you use the same anchor text (e.g. "women's running shoes") on different target URLs, you send a contradictory signal. Google, and you, no longer know which page should rank for that keyword. This is a typical keyword cannibalization problem that can be solved through the anchor texts of internal links: one target page per search intent, one clear anchor text per target.

Context vs. boilerplate: the Reasonable Surfer model. The idea of differentiating link weight by click probability goes back to a Google patent filed in 2004 (US7716225B1). The 2024 Google leak shows that Google does store features like anchor text font size. An attribute called droppedLocalAnchorCount suggests that some internal links are not counted (leak, weighting unknown). More in the section What the 2024 Google leak reveals about internal links. The idea: a link in body text that fits the paragraph's topic likely weighs more than a footer link.

Does Google confirm this? Google has made contradictory statements over the years, and controlled tests are lacking. Mueller said in 2025 that four identical sitewide links to the same page are "fine & common" (Google statement). Our advice: place the most important links in body text, where they help the reader. Footer links for support pages are fine. A link that fits the paragraph helps the reader anyway.

How many links per page? Google used to recommend an upper limit of 100 links per page. Google removed this recommendation. The current documentation says: "There's no magical ideal number of links a given page should contain." (Google statement) The Zyppy study shows a correlation between internal link count and traffic up to about 40 to 44 links, after which traffic drops (evidence, correlation). The authors attribute the decline to sitewide navigation: pages with 50+ links often have a large main navigation that dilutes the traffic effect.

Mueller clarified in 2025 that the warning about too many links applied to sites that link "from all pages to all pages". Four identical links to the same page are "fine & common" (Google statement).

Our takeaway: there is no ideal number. Link to what helps the reader on this page. But check whether the main navigation contains too many links that water down the structure.

Nofollow, sculpting and other myths

There are a few persistent rules around internal linking that are either outdated or wrong. Here are the four most important, with context:

Claim

Fact

Evidence strength

"No more than 100 links per page"

Google removed this recommendation. The current documentation names no ideal number.

Google statement

"Internal nofollow lets you redirect link equity to the remaining links"

Does not work. Mueller said in 2023: "You don't hoard anything when you make links nofollow." He called it "a common SEO myth". Since 2019, Google treats nofollow, sponsored and ugc as hints; the current documentation says: "Links marked with these rel attributes will generally not be followed."

Google statement

"The first link to a URL is the only one that counts"

Supported by several tests from 2008 to 2023, not confirmed by Google. Mueller said in 2018: "isn't something we have defined." Put the most descriptive anchor text in the first link.

Supported by tests (small tests)

"The title attribute of a link is a ranking factor"

There is no evidence for this in Google's documentation. Two German-language guides claim it anyway.

Unsubstantiated

Nofollow sculpting in detail: The idea of redistributing link equity through internal nofollow was a common SEO technique until 2009. Mueller called it in 2023 "a common SEO myth": "You don't hoard anything when you make links nofollow" (Google statement). How exactly Google handles link equity for nofollow links is not publicly documented. What is clear: the technique does not work. Google recommends robots.txt for internal steering, not nofollow (Google statement, documentation). The technique has been dead for 17 years. In our audits we still find it. Anyone using nofollow on internal links is hurting their own linking (experience).


The idea of distributing link equity according to the commercial value of a page is not new. Leslie Rohde described it in 2003 under the term "Profit Shaping" (Leslie Rohde, "Dynamic Linking, Volume I: PageRank Sculpting", 2010). His tools (nofollow tricks, JavaScript obfuscation, user-agent switching) are outdated. The thinking behind them is more relevant than ever.

The core question is: which pages generate the most revenue or have the highest search volume? And: do exactly those pages receive the most internal links? In many shops and portals the answer is no. Generic categories sit on enormous link equity (because they are linked in the main navigation), while brand and product pages with high demand barely receive any internal links.

Rohde also called this "Optimization by Subtraction": internal linking does not create new authority. It only redistributes the authority you already have. Without good content and backlinks the effect is limited. This honest framing belongs in every discussion about internal linking.

In May 2024, thousands of internal Google API documents became public. Rand Fishkin (SparkToro) and other SEO practitioners evaluated the documents. The original documents are publicly viewable as an auto-generated Elixir library (hexdocs, version 0.4.0). Google confirmed the authenticity of the documents to media outlets in late May 2024 (Search Engine Roundtable, 30 May 2024).

For internal linking, the leak is the first hard confirmation that Google stores link features as individual attributes. It shows that these data fields exist. It does not show how heavily they are weighted (leak, data up to August 2023).

The central attribute: onsiteProminence. The documentation describes it as: "Onsite prominence measures the importance of the document within its site. It is computed by propagating simulated traffic from the homepage and high craps click pages." Google calculates the importance of each page within its own site by propagating simulated traffic from the homepage and high-traffic pages through the link structure. Linking from those pages to a target page makes it more "prominent" (leak, attribute confirmed, weighting unknown).

Attributes relevant to internal links:

Group

Attribute

What Google stores according to the docs

What it means for internal links

Source

Link equity

homepagePagerankNs

Homepage PageRank, stored on every page of the site

The homepage is the equity source for the entire site

PerDocData

Link equity

sourceType

Quality of the linking page (High/Medium/Low Quality, correlates with index tier)

Links from strong pages on your own site likely carry more weight than links from thin pages

AnchorsAnchor

Link equity

topPrOnsiteAnchorCount

Counts onsite anchors from high-PageRank sources separately

Internal links from your highest-PageRank pages are counted as a separate, high-quality group

AnchorStatistics

Anchor counting

localAnchorCount / droppedLocalAnchorCount

Counts internal anchors and separately the dropped ones

Google counts internal links but drops some. Mass identical links stop adding value at some point

AnchorStatistics

Anchor counting

droppedRedundantAnchorCount / AnchorsRedundantAnchorInfo

Dropped per combination of domain and anchor text

Sitewide identical anchors do not scale linearly

AnchorStatistics, RedundantAnchorInfo

Anchor counting

parallelLinks

Number of additional links from the same source page to the same target domain

Multiple links are consolidated; five identical links are not five times the signal

AnchorsAnchor

Anchor content

text / origText

Anchor text stored normalized and in original form

Descriptive anchors give Google usable text about the target

AnchorsAnchor

Anchor content

fullLeftContext / fullRightContext

Text before and after the link stored

The surrounding sentence is part of the link signal. In-text links provide more context than bare link lists

AnchorsAnchor

Anchor content

fontsize

Font size of the anchor captured

Visual prominence is stored; whether and how it is weighted is unknown

AnchorsAnchor

Anchor content

anchorMismatchDemotion

Demotion when anchor text and target page do not match

Misleading anchors risk a demotion. "Click here" wastes signal

CompressedQualitySignals

Site structure

navDemotion

Demotion with navigation context (details undocumented)

Navigation is evaluated qualitatively. Use only as a hint

CompressedQualitySignals

Site structure

siteFocusScore

How focused a site is on a single topic

Topic clusters with tight internal linking support a clear topic assignment

TopicEmbeddings

Site structure

siteRadius

How far individual pages deviate thematically from the site core

Pages that deviate far from the core are measurable. Internal links should reflect topical proximity

TopicEmbeddings

Seven takeaways for internal linking:

  1. Homepage and high-traffic pages are the equity sources. onsiteProminence is calculated by propagating simulated traffic from the homepage and high-click pages. Important target pages should be reachable in few clicks from both.
  2. Choose link sources by strength. sourceType, pagerank of the source and topPrOnsiteAnchorCount show: Google records how strong the linking page is for every link. One link from your strongest guide page is worth more than ten links from thin archive pages.
  3. More links of the same kind eventually stop adding value. droppedLocalAnchorCount, droppedRedundantAnchorCount and parallelLinks confirm that Google drops internal and redundant anchors. A well-placed contextual link beats a sitewide identical link block.
  4. Choose anchor text that is descriptive and matches the target. anchorMismatchDemotion shows that the anchor is compared against the target. The anchor should say what the target page is about.
  5. The sentence around the link matters. fullLeftContext, fullRightContext and context2 store the text before and after the link. In-text links provide more context than bare link lists.
  6. Build topic clusters. siteFocusScore and siteRadius show that Google measures topic focus and how far individual pages deviate from the site core. A pillar page plus tightly linked subpages is exactly this structure.
  7. Visible links in the main content. fontsize shows that the visual appearance of a link is captured. Important links belong visibly in the main content, not small in the footer.

Limitations: The documentation describes data fields, not their weighting. Several fields are marked as DEPRECATED. The hexdocs pages are a third-party mirror, not an official Google source. How heavily any single field feeds into ranking cannot be inferred from the documents. Label in the rest of this guide: "leak (attribute confirmed, weighting unknown)".

Analyzing individual URLs is impractical at scale. We cluster by page type instead: homepage, main category, subcategory, filter/tag page, product/service page, guide/blog and utility pages (shipping, contact, legal notice). For each type we capture:

  • Internal link score (e.g. from Screaming Frog or Sitebulb) as a proxy for internal link equity
  • Average unique inlinks (how many different pages link to this type?)
  • GSC clicks and impressions as a proxy for demand
  • Revenue (if available, from analytics or the shop backend)

The sample data shows three imbalances that are typical in our audits (experience):

Utility pages hoard link equity without clicks. Shipping, contact and legal pages sit in the header or footer and receive almost as much link equity as main categories (in the example: link score 80, 5 clicks per month). They have almost no organic demand.

Guides and subcategories have demand but little link equity. Subcategories sit at link score 40 and 350 clicks in the example; guides at score 6 and 180 clicks. Both receive fewer internal links than their traffic warrants. Product pages (score 2, 60 clicks per page, but thousands of pages) show the same pattern at an even larger scale.

Filter pages drain link equity without return. Each filter URL picks up a few links through the faceted navigation. Across hundreds of pages, a significant amount of equity leaks away, at barely 20 clicks per page per month.

The most effective countermeasure is link boxes on main categories that point to high-demand subcategories and guides. This shifts equity to where demand is. Trimming the footer helps too: fewer links to utility pages gives the remaining footer links more weight. Filter pages with no search demand of their own should be linked non-crawlably or consolidated via canonical (see faceted navigation).

The link matrix documents, for each page type, which other page types are linked and through which element. It is the blueprint for your templates. An empty cell is either a deliberate decision or a gap you have not noticed yet.

Example: if category pages link to other categories via navigation but never to brand pages, even though brand pages drive the most traffic, that is where you start. You enter the planned link element in the cell (e.g. "C" for a content module with the top 10 brands of that category) and brief the development team accordingly.

Trimming the main navigation

The main navigation is a sitewide element and appears on every page. It therefore links with enormous equity. In many shops the mega menu contains hundreds of links, including filter combinations, third-level subcategories and login pages. All of these pages receive link equity whether they need it or not.

The countermeasure is a navigation diet: the navigation visible to crawlers is limited to the most important targets. The remaining links stay available for users but are loaded only on interaction (hover, click, scroll) and are therefore not visible in the initial HTML crawl. In one of our projects we reduced the crawlable navigation to around 200 links (experience, anonymized).

Important: Google renders JavaScript and may still see lazy-loaded links. The navigation diet works most reliably when the lazy-loaded links are absent from the initial HTML and do not appear automatically on render either. Honestly, the effect on PageRank distribution is hard to isolate. The measure does help with crawl steering, though, because Googlebot then discovers fewer irrelevant URLs.

Pyramids are the simplest form of deliberate linking: homepage links to categories, categories link to subcategories, subcategories link to detail pages. Breadcrumbs lead equity back up. This structure exists on almost every website. The lever is steering it deliberately: which categories sit in the main navigation (and therefore receive the most equity) and which are only reachable through link boxes or body-text links.

Loops form when detail pages send equity back to target categories or brands via breadcrumbs and contextual links (e.g. "Back to category", related products). Without these return links, equity flows into the depths and stays there. Rohde called this the "Home Page Push" in 2003 (Rohde, 2010). In practice, it is breadcrumbs and contextual return links.

Hub crosslinks are cross-references between silos that happen at hub level. A "shoes" category links to the "apparel" hub page, but an individual product page from "shoes" does not link directly to a product in "apparel". This keeps the topical clusters intact and equity is shared between the hubs that benefit most from it.

What no longer works

Three techniques found in older SEO books and guides have been ineffective or dangerous for years:

Nofollow sculpting (dead since 2009): internal nofollow does not redistribute equity, it destroys it. Details are above in the myths section.

JavaScript hiding for sculpting: hiding links from the crawler via DOM rewriting or onclick navigation does not work reliably since Google renders JavaScript with an evergreen version of Chromium. The Google documentation confirms: "Google Search runs JavaScript with an evergreen version of Chromium" (Google statement). Links without <a href> are still not recognized. The effect is therefore unpredictable: Google sees some hidden links, misses others.

User-agent cloaking: showing Googlebot a different link structure than the user is cloaking and violates the spam policies. Google detects this through rendering comparisons. The consequence can be removal from the index.


Part 4: Technical and special cases

Google can generally only crawl links that are <a> elements with an href attribute in the HTML or the rendered DOM. The Google documentation explicitly lists what does not work: routerLink without href, onclick navigation without a URL in href, javascript: URLs and <span> elements with href (Google statement). Our article on JavaScript SEO explains the technical details.

In practice we encounter these problems most often with single-page applications (SPAs), mobile navigations loaded via JavaScript, and filter elements that work via AJAX. The test is simple: does the link appear in the rendered HTML as <a href="/target/">? Screaming Frog can crawl in JavaScript rendering mode and shows the difference between the static HTML and the rendered DOM.

Faceted navigation and filters

Faceted navigation (colour, size, brand, price as combinable filters) is, according to Google, "by far the most common source of overcrawl issues site owners report to us", wrote Gary Illyes in December 2024 (Google statement). Filters generate thousands of URL combinations that Google must crawl but that rarely have their own search demand.

The Google documentation on faceted navigation recommends (Google statement):

Block facet URLs without search value via robots.txt. Alternatively, use URL fragments (#) which Google ignores, or set rel=canonical to the unfiltered main page. Nofollow on all filter links works less reliably in the long run because Google treats nofollow as a hint. A fixed filter order in the URL avoids duplicates. For empty or nonsensical combinations Google recommends a 404 status code.

For internal linking this means: filters that serve their own search demand (e.g. "women's running shoes" as a filtered result of a "running shoes" category) need their own indexable URLs with internal links. All other filter combinations should be invisible to the crawler. How to manage your crawl budget is covered separately.

Pagination after the end of rel=next/prev

Google has not used rel=next/prev since 2019. The current pagination documentation says: "Google no longer uses these tags, although these links may still be used by other search engines" (Google statement). Google recommends:

Paginated pages should be linked sequentially via <a href> (page 1 to page 2, etc.). Each page gets its own canonical URL (page 2 canonicalizes to itself). "Load more" buttons and infinite scroll need crawlable links so Google can reach the following pages. URL fragments (#) for page numbers do not work.

For internal linking this means: make sure pagination is built as real <a href> links. With "load more" via JavaScript, the most common mistake is that the crawler cannot reach subsequent pages because no crawlable link exists. The pagination article covers the technical implementation.

Finding and fixing orphan pages

Google writes: "Every page you care about should have a link from at least one other page on your site." (Google statement) A page without any incoming internal link (an "orphan page") is hard or impossible for Google to find, even if it is in the XML sitemap.

How to find orphan pages:

  1. Crawl your website with Screaming Frog or Sitebulb (mode: all internal URLs).
  2. Import your XML sitemap or GSC data into the crawler.
  3. Compare: which URLs are in the sitemap but have zero incoming internal links?
  4. Fix: place an internal link from at least one topically related page to the orphan page. More than one link is better.

The GSC shows the most linked pages and their sources in the "Internal Links" report (Google statement). The report is based on a sample, though, and shows a maximum of 1,000 rows (export up to 100,000).

Multilingual websites

On multilingual websites, internal linking in each language version is independent. The German version needs its own contextual links between the German pages, the English version between the English pages. The language switcher (e.g. a flag dropdown) links the language versions to each other and supplements the hreflang markup.

Three common mistakes:

  1. Language switcher without <a href>: If the language switcher works via JavaScript and has no crawlable link in the HTML, Google finds the other language versions only through hreflang and the sitemap. An <a href> in the HTML is safer.
  2. Internal links point to the wrong language version: A link in a German article that points to the English version of a page confuses users and sends an unclear signal. Always link to the page in the language you are currently writing in.
  3. hreflang without internal links: hreflang tells Google which pages belong together. Internal links tell Google which pages are important. Both are needed; neither replaces the other.

Ecommerce: category, product, guide

Ecommerce websites face a specific challenge with internal linking because they manage thousands of product and category pages across multiple hierarchies.

Category pages are the most important distributors. They sit in the navigation and link to subcategories, brands and products. Check whether the links on the category page lead to the pages that drive the most revenue. Link boxes like "top brands" or "popular products" on category pages are a simple lever.

Product detail pages (PDPs) often receive only a single internal link (from the category listing) and return equity through "similar products" and breadcrumbs. For PDPs with particularly high search volume (e.g. well-known products or bestsellers), it is worth placing additional links from guides and category pages.

Guides and blogs typically link to categories and products, providing topical relevance and link equity. The most common mistake: guide articles are published but only link generically to the homepage, or not to product pages at all.


Part 5: AI search, audit and measurement

AI search engines like ChatGPT, Perplexity and Google AI Overviews are becoming a relevant traffic source for many websites. How important is internal linking for visibility in these systems? Here is where things stand (November 2026):

Evidenced: Most major AI crawlers do not render JavaScript. A Vercel/MERJ analysis from December 2024 shows: "none of the major AI crawlers currently render JavaScript" (evidence, vendor network data, Vercel network only). Affected are GPTBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot and others. Only Google's infrastructure (Gemini) and AppleBot render. The consequence: navigation and links must be in the HTML server-side for AI crawlers. A pure JavaScript navigation is invisible to these crawlers. Our article on understanding AI crawlers and bots and the article on rendering for AI crawlers explain the technical details.

Evidenced: Google says about its own AI search (AI Overviews, AI Mode): "no additional requirements" beyond the normal SEO best practices (Google statement). The recommendation: make content "easily findable through internal links". Google's AI optimization guide from May 2026 confirms this line: content must be crawlable, JavaScript pages need standard JS SEO best practices, and there is no requirement to break content into small pieces (Google statement). The guide does not mention internal links specifically.

First correlation: A study (May 2026, 8,500 keywords) shows that pages in AI Overviews had more internal links on average (evidence, correlation, one market, one week). The authors warn that the exact magnitude is "the most cohort-sensitive number in the report". Causal proof is missing.

Opinion only: The claim that internal linking improves citations in ChatGPT and other LLMs circulates in many blogs. We found no causal evidence for it in our research.

Calculating internal PageRank

How do you find out how link equity is distributed across your website? Two crawlers offer a proxy:

Screaming Frog Link Score: A metric between 0 and 100 that calculates the relative value of a page based on its internal links (tool metric, not a Google signal). You find it after the crawl under Crawl Analysis > Link Score. Our Screaming Frog article explains the setup.

Sitebulb URL Rank (UR): Sitebulb calls its variant a "proxy for 'internal PageRank'" (tool metric).

Both values are approximations. They say nothing about how Google weights things internally. But they reliably show where equity goes and where it does not. For the analysis by page type (see above), export the link score and calculate the average per page type. Alongside clicks and impressions from the GSC, this gives you a clear picture of the imbalance.

Doing your own calculation: If you want exact numbers, you can calculate internal PageRank from the crawl export yourself. The formula is the random surfer model (background in What is PageRank?): every page starts with a base value, passes the damping factor (typically d = 0.85) of its value evenly across its outgoing links, and the remainder (1 minus d, i.e. 0.15) is distributed evenly across all pages. After 20 to 40 iterations the model converges. In practice the crawler's link score is enough, but for large projects with several hundred thousand URLs a custom calculation (e.g. in Python or R) can be more precise and flexible.

An internal linking audit consists of seven steps. The tools are interchangeable; the method matters.

1. Google Search Console: open the links report. Under "Links" > "Internal Links" you see the pages with the most internal links. Compare this list with your most important pages. If your top revenue pages are missing, the priorities are off. Note: the report is based on a sample and does not fully represent the linking (Google statement).

2. Crawl. Crawl the entire website with Screaming Frog, Sitebulb or a comparable tool. Enable JavaScript rendering if your website uses JavaScript for navigation.

3. Find orphan pages. Import the XML sitemap and GSC data into the crawler. Every URL in the sitemap with zero incoming internal links is an orphan page.

4. Calculate link score. Run the crawl analysis (Screaming Frog: Crawl > Crawl Analysis) and export the link score. Cluster by page type and compare with traffic.

5. Analyze anchor texts. Export all internal anchor texts (Bulk Export > Links > All Inlinks). Look for generic texts, empty anchors and identical anchor texts pointing to different URLs.

6. Check click depth. The crawler shows the click depth of every page. Sort by depth and check whether important pages sit too deep (more than four to five clicks).

7. Prioritize findings. Sort the findings by impact: orphan pages with search volume first, then anchor texts of your highest-revenue pages, then click depth, then broken links.

Ahrefs and Semrush offer additional automation: Ahrefs shows "Internal link opportunities" (pages that mention a keyword but do not link to it). Semrush flags nine error patterns in its site audit for internal linking (broken links, nofollow, orphan pages, click depth, redirect chains).

Log files show how Googlebot and AI bots actually crawl your website: which pages are requested how often, and which are never visited. This is the only source that shows what the crawler does. Our article on log file analysis explains the process.

Measuring impact

After changing the internal linking you need a method to detect the effect.

Before/after in the GSC: Export clicks and impressions of the affected pages before the change (e.g. four weeks) and after the change (four to eight weeks waiting time so Google crawls the new links and the change feeds into rankings). Compare the totals and watch for seasonal fluctuations.

Crawl comparison: Crawl before and after the change and compare link score, click depth and the number of orphan pages. Tools like Screaming Frog allow crawl comparisons.

GSC indexing report: Check after the change whether pages that were previously not indexed are now indexed. The "Pages" report shows the status.

How fast do new links take effect? Gary Illyes showed at Search Central Live Barcelona in October 2026 that Google typically picks up changes to internal links ("link annotations") within minutes to one to three weeks; in the worst case it takes months (Google statement, per attendee report, primary slides unverified). Four to eight weeks of waiting time before measuring covers this range.

A clean causal proof is difficult with internal linking changes because other factors change at the same time (algorithm updates, seasonality, new content). Therefore: change only the linking where possible and keep other variables constant. Document the change with its date and scope.

Quick wins and checklist

The following measures deliver the fastest results in our experience (experience):

  1. [ ] Find orphan pages and link them from at least one topically related page
  2. [ ] Let pages with many backlinks (high external authority) link to priority pages
  3. [ ] Retroactively link older articles to new content
  4. [ ] Fix broken internal links and redirect chains to their final target
  5. [ ] Bring priority pages to a maximum of three to four clicks depth
  6. [ ] Replace generic anchor texts ("here", "more") with descriptive ones
  7. [ ] Check whether the same anchor text points to different target URLs (cannibalization signal)

FAQ

What is the difference between internal and external links?
An internal link points to a page on the same domain. An external link points to a different domain. A backlink is an external link from the perspective of the target site: when google.com links to buzzmatic.com, that is a backlink for buzzmatic.com.

How many internal links per page are ideal?
There is no ideal number (Google statement). Link to what helps the reader on this page. Watch out for a main navigation that dilutes the link structure.

Is internal linking a ranking factor?
Google confirms that internal links are a signal for the importance of a page: "Every page you care about should have a link from at least one other page on your site" (Google statement). Mueller called them "one of the biggest things that you can do on a website" in 2022. There is no direct ranking-factor statement, but the effect through crawling, link equity and anchor texts is evidenced.

Do internal links improve visibility in ChatGPT and Perplexity?
There is no causal evidence. A study (May 2026, 8,500 keywords) shows a correlation for Google AI Overviews: pages with more internal links appear more frequently in AI Overviews (evidence, correlation, one market, one week). What is evidenced: most AI crawlers do not render JavaScript, so links must be in the HTML. For Google AI Overviews the normal SEO best practices apply, "no additional requirements" (Google statement).

Should I use nofollow on internal links?
In almost all cases, no. Mueller called internal nofollow in 2023 "a common SEO myth": "You don't hoard anything when you make links nofollow" (Google statement). For internal steering, Google recommends robots.txt. Exception: login pages or similar URLs that should never be indexed, where nofollow serves as a supplementary hint at best.

What do I do with pages that should not be link targets (legal notice, privacy policy)?
Link them in the footer. They need internal links because they are trust signals. But you do not need to link them in the main navigation or body text.

How often should I review internal linking?
After every major change to the site structure (migration, new template, new category) and at least once per quarter on active websites.

Portrait of Raoul Schreck, CEO & Co-Founder of Buzzmatic
Raoul Schreck

CEO & Co-Founder

Raoul has been doing SEO for over 15 years, starting at Pixum, then Groupon and Fashion For Home, before co-founding Buzzmatic with Oguzhan in 2010. Today he drives BuzzView, the GEO discipline and consulting for our enterprise clients.

Comments

No comments yet. Write the first one!

Write a comment

Your take, questions or additions. We read everything.

By submitting you agree that we publish your name and comment. Your email stays private.

Need support?

Our experts are happy to help.

Get in touch