Within the structure of modern search engine optimization, internal link (Internal Link) often reduced to manual actions lacking quantitative basis: installing plugins that automatically insert exact keywords, scattering links in the footer, or building huge mega-menus in an attempt to “flatten” the website. However, through the lens of computer science and Google's actual technical documentation, the internal linking network is a Weighted Directed Graph operating under the influence of a series of complex machine learning algorithms.

This in-depth research article by DPS.MEDIA comprehensive analysis of how internal links operate, based on three solid scientific pillars of evidence: the patents granted to Google LLC, more than 2,500 pages of leaked technical documents from Google Content API Warehouse (May 2024), and official algorithm documents from the Google Search Central engineering team.

1. Reasonable Surfer Model (Patent US7716225B1): PageRank Distribution Mechanism Based on Click Probability

In 1998, Larry Page and Sergey Brin introduced the algorithm PageRank original version based on assumptions Random Surfer Model (random web surfer model). In this model, every link appearing on a web page carries equal PageRank transmission value: if a page has 10 links, each link receives exactly 1/10 (10%) of the value flow (Link Equity), regardless of whether the link sits at the top of the article, in the menu bar, or deep in the footer.

However, this unrealistic assumption has been completely eliminated by Google since the patent was filed and implemented US7716225B1 — Ranking documents based on user behavior and/or feature data (the authors include Google's legendary engineers Jeffrey Dean, Corin Anderson and Alexis Battle). This patent gave rise to Reasonable Surfer Model (Reasonable web surfer model).

According to the technical analysis by search algorithm expert Bill Slawski in the document Google’s Reasonable Surfer Patent Updated on SEO by the Sea, Reasonable Surfer calculates actual click probability (Click Probability) of user interaction with each individual link, based on the visual attributes of the HTML DOM:

  • Physical position in the page layout (Location in DOM): The link sits in the first third of the main content body, where it has a 5 to 8 times higher chance of being clicked than links pushed to the bottom of the page or placed in the author box.
  • Visual Features: Font size, font weight, and color contrast relative to the surrounding text directly affect the transmission weighting coefficient.
  • Boilerplate Area Eliminated: Links located in the footer, sidebar, disclaimer, or repeated navigation bars (boilerplate links) have a probability of users actively reading and clicking them approaching zero, meaning the PageRank weight passed through these positions is almost entirely neutralized.

Experimental Conclusion: The placement of a link determines up to 80% of the PageRank value passed on. An internal link placed naturally within the introductory paragraph at the beginning of an article delivers superior ranking value compared to a list of 10 links crudely clustered at the bottom of the page.

2. Link Index Tiers (3-Tier Link Index) & Actual Click Metrics from the 2024 Google API Leak

In May 2024, a historic event in the search industry took place when the internal document repository Google Content API Warehouse was publicly leaked, and was published and analyzed by Rand Fishkin (SparkToro) and Mike King (iPullRank) in their report An Anonymous Source Shared Thousands of Leaked Google Search API Documents. This document reveals previously unpublished details about how Google stores, evaluates, and removes internal links.

Google's link processing system does not store every link on a single data plane but separates them into three quality tiers (3-Tier Link Index):

  1. High-Quality Tier (high-quality tier): Reserved for links originating from pages with high actual engagement that are indexed by Google immediately and pass full PageRank value as well as Anchor Equity.
  2. Medium-Quality Tier (Middle Tier): Collected and updated periodically, it passes on only part of the ranking value.
  3. Low-Quality Tier: Documents of poor quality or with no user interaction at all. Links within this tier are isolated, pass no value to the destination page, and may be removed entirely from the link graph memory.

In particular, the internal source code confirms the existence of the module AnchorsAnchor with the attribute SourceType and the document metadata management module RepositoryWebrefDocumentMetadata field tracking TotalClicks. When a website records no views or clicks from the Chrome browser and Google Search (TotalClicks = 0), that page is flagged as a “dead document” (Dead Document). Any internal link created from this page is demoted to the Low-Tier level and ignored by Googlebot.

At the same time, the ranking system based on user behavior is called NavBoost continuously monitors click signals including goodClicks (clicks followed by a long on-site dwell time), badClicks (clicks away from the page immediately) and lastLongestClicks (the final click that fully satisfies the user's need). An internal link that prompts users to click and continue engaging deeply on the destination page sends a very strong reinforcing signal to NavBoost to raise the rankings of the entire topic cluster.

3. First Link Priority & Single Link Rule: Googlebot's Parsing Mechanism

A classic mistake many editorial teams make is inserting multiple links that point to the same destination URL within one article (for example, one link in the breadcrumb menu at the top and another link containing the strategic keyword in the middle of the body). From the perspective of the DOM Parser algorithm, this behavior triggers the rule First Link Priority (Priority given to the first link).

As analyzed in detail on Search Engine Journal through the commentary of search expert John Mueller in the article Google's Mueller On First Link Priority & Link Obfuscation, when Googlebot scans the HTML source code from top to bottom and detects two or more tags that all lead to a single destination URL:

  • Googlebot only records Anchor Text of the first link appears within the HTML DOM hierarchy tree to analyze semantics and pass ranking signals.
  • All anchor text of subsequent links in the body pointing to that same URL is completely stripped of semantic value by the parser.

Many developers try to "get around" this rule with link obfuscation techniques (Link Obfuscation), such as using JavaScript events onClick or tag encoding . John Mueller clearly warned that this technique breaks standard HTML semantic structure, preventing Googlebot from understanding the link hierarchy tree and potentially being judged as search engine manipulation.

Single Link Rule: Within each article, each strategic target URL may only appear exactly 1 time only in the main content (Main Body), with the anchor text composed most precisely so that Google fully recognizes the semantic value.

4. Information Gain (Patent US10296577B2): Building Links That Serve as a "Logical Next Step"

Google is increasingly tightening its standards for evaluating duplicate content and thin content (Thin Content). The focus of this shift lies in the patent US10296577B2 — Contextual Estimation Of Link Information Gain (announced in a 2024 update). This algorithm introduces the concept Information Gain Score (Diem gia tang thong tin).

According to the analysis from the search algorithm research platform in the document What is Information Gain? on Search OS, when a user browses through web pages in a search journey (Search Journey), Google estimates the amount of new information a document can add beyond what the user has already read:

  • If an internal link leads readers to an article that merely rehashes or repeats the exact same definitions they just read in the source article, the Information Gain Score of the target page will be 0.
  • Users immediately press the back button (pogo-sticking), which the NavBoost system flags as badClicks, leading to a decline in the rankings of both the source page and the target page.

Therefore, a scientifically sound internal link architecture must always follow the principle Next Logical Step (logical next step) in the reader's problem-solving path. The standardized 4-step journey model consists of:

Foundational Concepts & Core Context ➔ Detailed Technical Execution Process ➔ Exception Handling, Risks & Real-World Case Studies ➔ Packaged Technology Solution / Service.

Every correct click unlocks clear added value, turning the website into a closed knowledge ecosystem that keeps users fully achieving their goals without returning to the Google search results page (SERP).

5. Diversifying 3-Tier Anchor Text & Manipulative Indexing Filter Mechanisms IndexingDocjoinerAnchorSpamInfo

One of the most common mistakes in traditional SEO optimization is using 100% exact match keywords as anchor text from dozens of satellite articles linking to the category page or core service page.

Patent US10068022B2 — Identifying topical entities of Google engineers (Haran Pilpel, Ziv Bar-Yossef) demonstrated that Google recognizes entities based on the network of semantic phrases surrounding a link, rather than relying solely on single keywords. In particular, the Google Warehouse Leak 2024 documents exposed the module IndexingDocjoinerAnchorSpamInfo.

This module is responsible for analyzing the distribution spectrum of anchor text pointing to a document. If an unnatural repeating pattern (Unnatural Anchor Pattern) is detected, such as anchor text that is overly uniform in its vocabulary, the spam filter will trigger a ranking suppression mechanism (Demotion Filter), completely eliminating the ranking-boosting power of the entire link network.

To optimize according to scientific standards, the anchor text distribution ratio should be designed in balance using a 3-tier formula:

Anchor Text Layer Optimization Ratio Nature & Illustrative Example
1. Exact Topical Entity
(Core entities)
25% Clearly affirm the exact entity name (Example: Modern customers expect experiences tailored to their needs and preferences at every touchpoint along the journey. Marketing automation allows you to:, Google Search Console).
2. Partial Intent Match
(Expanded Search Intent)
45% Tied to the reader's specific action or question (For example: sustainable website structure optimization solution, internal link audit process).
3. Semantic Phrase / LSI
(Natural contextual phrase)
30% Smooth transitional phrases tied to the context of academic sentences (Example: document analyzing link patterns, technical report on PageRank distribution).

6. Click Depth vs URL Depth: Freeing Your Thinking from the Illusion of the “3-Click Rule”

For many years, a common misconception was that the shorter the URL structure, the fewer slashes it contains / the higher a website ranks (URL Depth). To counter this, many webmasters have crammed hundreds of links into the mega-menu to turn every subpage into a “1 click from the homepage” distance.

Search science completely rejects this understanding. Technical research from Unveil SEO in the article Crawl Depth and the Three-Click Rule Google Never Made together with John Mueller's official statement recorded by Search Engine Journal in Google: Click Depth Matters for SEO, Not URL Structure has definitively confirmed:

“Googlebot does not care at all about folder structure or the number of slashes in a URL path. The only factor that carries meaning for ranking and indexing is Click Depth (Depth of Clicks) — that is, the minimum number of clicks needed to get from the Homepage to the destination page through meaningful links.”

At the same time, according to the official technical guidelines from Google Search Central about Large Site Crawl Budget Management (Managing Crawl Budget for Large Websites), abusing Mega-menus actually creates a negative effect: it "raises the floor but lowers the ceiling." Mega-menus spread the homepage's PageRank evenly across hundreds of secondary pages, weakening the flow of authority to the strategic conversion pages.

The only optimal solution is to use Contextual In-content Links (contextual links within the article body) to create "express bridges," bring service pages or in-depth analysis articles closer to the homepage within the network graph structure.

7. Content Silo & Topic Authority: Protecting Notability Score and NSR Signals

May 2023, Google officially published documentation on its topical authority system through a blog post Understanding news topic authority on Google Search Central Blog. This system measures the level of expertise and trustworthiness of a website within a narrow field of specialization.

In the 2024 API Warehouse leak documents, the algorithms that measure authority, such as the score NSR (Normalized Site Result) and the index Notability is directly affected by the integrity of the topic cluster. When internal links are designed chaotically, for example an article about SEO techniques placing random cross-links to an article about travel or nutrition, the semantic signals of the entire website will be fragmented, severely reducing its expertise score in the eyes of machine learning.

Model Content Silo (isolated content silo) requirements:

  • Every article within a topic cluster (Cluster) must be tightly interlinked and point back to the pillar page (Pillar Page) to reinforce the central entity.
  • Absolutely avoid lateral links (Lateral Link) to topics with no semantic relevance.
  • When commercial navigation is needed, Cluster pages may only direct traffic to service pages that belong to the same specialist ecosystem.

To comprehensively optimize business performance from paid-free search campaigns, businesses can refer to comprehensive SEO services by DPS.MEDIA — a solution combining a scientifically structured link graph architecture, semantic entity optimization, and an algorithmically sound Content Silo system.

8. Technical Cross-Reference Matrix & 6-Point Pre-Flight Checklist

To turn the entire theoretical foundation and patents mentioned above into standard operating procedures (SOPs), the matrix below systematizes the key principles:

Algorithm / Patent Core Measurement Mechanism Common Mistakes Practical Implementation Standards
Reasonable Surfer
(Patent US7716225B1)
Click probability based on position and visual display attributes. Stuffing links into the footer, sidebar, author box, or grouping links at the end of the article. Place the link at First 1/3 of the body, within a standalone paragraph with clear contrast.
3-Tier Link Index
(Google Warehouse Leak)
Classifying links based on real interaction data (TotalClicks). Inserting links on old articles with 0 views (“dead content”). Only select articles that currently have real traffic and clicks as link-source hubs.
First Link Priority
(DOM Parser Rule)
Only the anchor text of the first link tag pointing to the same URL is counted. Place 2-3 links pointing to the same destination URL within the article. Strict Compliance Single Link Rule: Each URL appears exactly 1 time.
Information Gain
(Patent US10296577B2)
Measures the incremental amount of information within the user journey. Linking to a page that merely repeats old knowledge causes Pogo-sticking. Link design plays the role of Next Logical Step in the problem-solving journey.
Anchor Spam Filter
(Docjoiner Module)
Filter for detecting precise anchor-text density manipulation. Uses 100% exact-match keywords pointing to the category page. Scientific allocation: 25% Core Entity, 45% Expansion Intent, 30% Natural Context.
Click Depth Optimization
(Crawl Budget)
Minimum number of clicks from the homepage to the target document. Overuse of mega-menus flattens the structure and dilutes PageRank. Use contextual links within the article to shorten the click distance.

Before publishing any article to the system, the editorial team needs to cross-check 6-Point Review Checklist (Pre-flight Checklist):

  • Inappropriate budget allocation and strategy 1. Location Anchor Point: Has the most important link, pointing to the Pillar page, been fully placed within the first third of the main content body?
  • Inappropriate budget allocation and strategy 2. URL Uniqueness: Are any URLs repeated 2 or more times in the article's source code?
  • Inappropriate budget allocation and strategy 3. Silo Integrity: Does the link mistakenly point to a topic group that is semantically incompatible in terms of entity meaning?
  • Inappropriate budget allocation and strategy 4. Information added value: Does the destination page provide substantive follow-on knowledge for readers, or does it merely repeat old content?
  • Inappropriate budget allocation and strategy 5. Natural Anchor Text Ratio: Do the anchor texts avoid meaningless phrases ("click here", "read more") and avoid repeating the exact main keyword 100% across other articles?
  • Inappropriate budget allocation and strategy 6. Source Page Strength: Is the page carrying the link a document that is receiving actual traffic and real user engagement?

9. Academic References

  1. Dean, J., Anderson, C., & Battle, A. (Google LLC). Ranking documents based on user behavior and/or feature data. U.S. Patent No. US7716225B1. U.S. Patent and Trademark Office.
  2. Slawski, B. (2016). Google's Reasonable Surfer Patent Updated. SEO by the Sea. https://www.seobythesea.com/2016/04/googles-reasonable-surfer-patent-updated/
  3. Fishkin, R., & King, M. (2024). An Anonymous Source Shared Thousands of Leaked Google Search API Documents. SparkToro & iPullRank. https://sparktoro.com/blog/an-anonymous-source-shared-thousands-of-leaked-google-search-api-documents-with-me-everyone-in-seo-should-see-them/
  4. Southern, M. G. (2024). Google's Mueller On First Link Priority & Link Obfuscation. Search Engine Journal. https://www.searchenginejournal.com/googles-mueller-on-first-link-priority-link-obfuscation/582316/
  5. Google LLC. Contextual Estimation Of Link Information Gain. U.S. Patent No. US10296577B2. U.S. Patent and Trademark Office.
  6. Search OS Research Team. What is Information Gain? Algorithmic Breakdown. Search OS. https://searchos.io/knowledge/information-gain
  7. Pilpel, H., & Bar-Yossef, Z. (Google LLC). Identifying topical entities. U.S. Patent No. US10068022B2. U.S. Patent and Trademark Office.
  8. Unveil SEO Technical Team. (2023). Crawl Depth and the Three-Click Rule Google Never Made. Unveil SEO. https://unveilseo.com/blog/crawl-depth
  9. Schwartz, B. (2018). Google: Click Depth Matters for SEO, Not URL Structure. Search Engine Journal. https://www.searchenginejournal.com/google-click-depth-matters-seo-url-structure/256779/
  10. Google Search Central Documentation. Large Site Crawl Budget Management. Google for Developers. https://developers.google.com/search/docs/crawling-indexing/large-site-managing-crawl-budget
  11. Google Search Central Blog. (2023). Understanding news topic authority. Google for Developers. https://developers.google.com/search/blog/2023/05/understanding-news-topic-authority?hl=vi

Research & Solution Consulting Unit Information: DPS.MEDIA JOINT STOCK COMPANY

DPS.MEDIA is a pioneer in applying data science, artificial intelligence (AI), and advanced algorithms to Digital Marketing strategic management and sustainable website ranking development for SMEs.

  • Should change IP if listed > 2 times Address: 56 Nguyen Dinh Chieu, Tan Dinh Ward, Ho Chi Minh City
  • 📞 Hotline / Zalo: 0961 545 445
  • ✉ Email: marketing@dps.media
  • 🌐 Official website: https://dps.media
  • 🏢 Tax identification number: 0318700500

Ý kiến bạn đọc (8)

DPS SECURITY

Đăng nhập để bình luận

Đăng nhập nhanh 1-chạm bằng Google để chia sẻ ý kiến của bạn về bài viết.

Nguyễn Minh Đức
Nguyễn Minh Đức Experts 20/09/2026

If a site has several thousand URLs, should internal links be prioritized based on actual clicks, or should the silo structure by topic group be kept as is? We previously built quite a rigid silo structure, but some pages still receive almost no traffic.

Linh Nguyen
Linh Nguyen 20/09/2026

In my view, you should not choose one over the other. Silos help Google understand topical relationships, while internal links should prioritize paths that are genuinely useful for readers to keep going. Some URLs sit in the same silo, but if they are not contextually related, linking to them won't help much.

vuducnhuong.ndi
vuducnhuong.ndi 20/09/2026

Right, I once cut a lot of links from the sidebar and footer and moved to contextual in-content links, and navigation behavior changed noticeably. But you still need to monitor data for each site and not treat one model as a fixed formula.

Quang Bùi
Quang Bùi Experts 20/09/2026

The First Link Priority part caught my attention. If a URL is already linked in the menu and then another anchor appears in the article, does Google really consider only the first anchor, or does it still depend on how the system understands context?

thanhdiachatk39
thanhdiachatk39 20/09/2026

It should be understood as a signal-processing issue rather than a simple rule like 'the first link wins absolutely.' The menu, link placement, anchor text, and context can all be analyzed by the system, so cramming multiple anchors for the same URL is not an optimal approach.

septemberflower
septemberflower 20/09/2026

My favorite part is the distinction between Click Depth and URL Depth. Some sites have URLs that look deep, yet users still reach content very quickly thanks to good navigation, so counting slashes in a URL to evaluate structure is somewhat mechanical.

Huỳnh Tuấn Kiệt
Huỳnh Tuấn Kiệt 20/09/2026

The anchor text diversification section is quite practical. We once inserted too many exact-match anchors automatically, then had to review them because the articles read very awkwardly and many links added no extra information.

Quê Hà Nội
Quê Hà Nội 20/09/2026

One point I would like to ask about is the cost of auditing internal links for a large website. If there are tens of thousands of URLs, doing it manually is likely not feasible, so it probably requires combining data crawling with contextual review of each page group.

Ready to scale your revenue with DPS.MEDIA?

Partnering with 5,400+ SMEs to optimize marketing costs, elevate brand authority, and automate growth with 2026 AI technology.