Internal Linking: Build Clear Paths for People and Search Engines

Direct answer

Internal linking is the practice of connecting pages on the same site with ordinary crawlable links. A useful internal link tells a reader what they can do next, gives the destination a descriptive relationship to the source, and helps search engines discover and understand the site. The practical rule is simple: link when the destination is the best next step for a real user, use anchor text that describes that destination, and make important pages reachable through stable navigation or contextual paths.

There is no official ideal number of internal links per page, no safe keyword-anchor percentage, and no public formula for how much authority a link passes. Ten useful links can be better than three vague ones or one hundred templated ones. Architecture, context, click depth, crawlability, destination quality, and maintenance all matter.

Start with page purpose. Give every search task one canonical owner, connect prerequisites to procedures and procedures to outcomes, then audit for orphan pages, broken destinations, redirects, vague anchors, and competing owners. Internal linking is not a decoration added after publishing. It is the navigation system of your knowledge base.

What you will be able to do

By the end, you will be able to:

• distinguish navigational, contextual, breadcrumb, and utility links;
• create links that Google can crawl and people can understand;
• select descriptive anchors without forcing exact-match phrases;
• design hubs and spokes around user journeys;
• connect prerequisites, lessons, tools, and next steps;
• measure click depth without worshipping an arbitrary threshold;
• find orphan pages and false orphans;
• repair broken, redirected, canonicalized, and duplicated destinations;
• review template links without turning the footer into a keyword cloud;
• monitor the internal link graph as the site changes.

What an internal link actually does

An internal link is an HTML link whose source and destination belong to the same site. It can perform four jobs at once:

1. Navigation: it helps a person move toward an answer or action.
2. Discovery: it gives crawlers a path to a URL.
3. Context: its surrounding words and anchor describe the relationship between two pages.
4. Architecture: repeated paths reveal hierarchy, sequence, and canonical ownership.

These jobs overlap, but they are not identical. A sitemap can support discovery without creating a good learning path. A breadcrumb can expose hierarchy without explaining why two lessons are related. A contextual link can clarify a concept without belonging in global navigation.

Google's link best practices say links are generally crawlable when they use an a element with an href. That sounds basic because it is basic. A styled div, a click handler, or a script that only creates a destination after interaction may work for a user in one browser while failing to provide a dependable crawl path.

Use progressive enhancement. The destination should exist in the HTML link. JavaScript may add behavior, tracking, or interface polish, but it should not be the only place where the route exists.

The four internal-link layers

Primary navigation

Primary navigation carries durable choices such as Product, Academy, Guides, Glossary, and Free tools. Keep it small enough to scan. It should reflect the site's principal jobs, not every query you want to rank for.

Hubs and collections

A hub helps a user choose among closely related destinations. An Academy hub can order lessons by skill level. A comparison hub can group software evaluations. A glossary can expose definitions alphabetically and by topic.

A hub needs its own purpose. If it merely repeats the first paragraph from every child page, it becomes a thin index. Add selection guidance, prerequisites, sequence, and the distinctions a reader needs to choose.

Contextual links

Contextual links sit inside the explanation. They are often the clearest way to express relationships:

• prerequisite to procedure;
• definition to application;
• problem to diagnostic;
• claim to evidence;
• guide to template;
• comparison to product capability;
• lesson to the next lesson.

Place the link where the need becomes apparent. A generic related-posts block at the end cannot replace a relevant link beside the concept.

Breadcrumbs

Breadcrumbs show position in a hierarchy and offer a route upward. Google's breadcrumb documentation recommends representing a typical user path rather than mechanically copying the URL structure. Visible breadcrumbs and BreadcrumbList markup should describe the same real hierarchy.

Markup does not replace visible navigation. It also does not guarantee a rich result. Use it because the relationship is true and useful.

Anchor text: describe the destination

Good anchor text answers, "What will I get if I follow this link?"

Weak: click here
Better: download the internal link audit template

Weak: learn more
Better: learn how canonical URLs resolve duplicates

Weak: SEO
Better: technical SEO crawlability checklist

Descriptive does not mean repeating an exact keyword unnaturally in every link. Use language that fits the sentence and accurately represents the destination. Natural variation follows from different source contexts. Do not create an anchor-ratio target and then rewrite useful prose to satisfy it.

The surrounding sentence matters too. "Before changing canonicals, reconcile indexability signals" creates a stronger relationship than a keyword inserted into an unrelated paragraph.

Images as links

When an image is the link, useful alternative text helps describe the destination and the image's function. Avoid empty or stuffed alt text. If the image is decorative and a nearby text link performs the action, keep the accessibility behavior intentional.

How many internal links should a page have?

As many as the page needs to complete its task and offer sensible next steps, but no more than can be maintained and understood. Google does not publish a target such as five links per thousand words.

Instead of counting links alone, ask:

• Does each link serve a current user need?
• Is the destination the canonical owner of that task?
• Are important actions visible without searching the footer?
• Are repeated template links useful across every page using the template?
• Does the page feel interrupted by links that add no value?
• Can each destination be maintained?

Use link counts as anomaly detection. A lesson with zero contextual links may be isolated. A 1,500-word article with 180 commercial anchors may be overlinked. Neither number proves the verdict by itself.

Build paths from tasks, not keywords

Start with the route-and-intent map from Step 15. Every page already has a primary task and canonical owner. Now add relationships.

For each page, define:

1. what the reader should know first;
2. what they are trying to complete now;
3. which evidence or tool helps them complete it;
4. what decision follows;
5. where they go if this page is not the right fit.

That produces a directed graph rather than an all-to-all mesh. A beginner lesson can link forward. An advanced lesson can link back to prerequisites. A tool can link to its instructions. A commercial page can link to proof and constraints. Each direction has a reason.

Click depth without mythology

Click depth is the minimum number of link selections needed to reach a page from a chosen start, usually the home page or a section hub. It is useful because buried pages can be harder to discover and harder for users to find.

There is no universal rule that every URL must be within exactly three clicks. A small conversion page and a long-tail archive do not have equal importance. Measure depth from multiple meaningful entry points and prioritize by page role.

The goal is not a flat site. It is a comprehensible site where important destinations have deliberate paths.

Find orphan pages correctly

An orphan page has no discoverable internal link from the crawl set you are evaluating. "Orphan" is relative to the dataset. A crawler may call a URL orphan because it started in the wrong subdomain, could not render a menu, lacked authentication, or did not receive a URL list.

Reconcile at least four sources:

• a crawl from representative entry points;
• XML sitemaps;
• analytics landing pages;
• Search Console or server-log URLs.

Then classify the URL:

• Link: valuable and intentionally indexable, but missing a path.
• Redirect: a retired duplicate with a clear replacement.
• Noindex or restrict: useful to users but not a search landing page.
• Remove: obsolete, empty, unsafe, or created by error.
• Investigate: unknown owner, tenant, or data state.

Do not automatically link every discovered URL. Some pages should not exist or should not be indexable.

Audit destinations before adding links

An internal link is only as useful as its destination. Before adding or automating a link, check:

• HTTP status;
• final destination after redirects;
• canonical target;
• indexability;
• locale;
• content completeness;
• task ownership;
• tenant and access boundaries;
• update status.

Link directly to the final canonical URL when you control both pages. Redirect chains waste time and hide stale architecture. A canonical pointing elsewhere can mean the link is reinforcing the wrong owner.

Ahrefs examined links pointing to a sampled set of 2,062,173 websites from a January 2013 crawl and reported that 66.5% had rotted after nine years. That is long-horizon vendor research across links, not a universal current internal-link failure rate. The useful lesson is maintenance: destinations disappear, redirect, or change purpose. Recheck them.

The internal link audit table

Use the downloadable Internal Link Graph Auditor. One row represents one source-to-destination relationship.

At minimum, produce these reports:

1. orphan destinations;
2. broken destinations;
3. redirected destinations;
4. vague or duplicated anchors;
5. conflicting canonical owners;
6. prerequisite-order gaps;
7. high-depth important pages;
8. template links with low relevance.

A graph view is useful for seeing isolated components and over-centralized templates. The row-level table remains the operational truth because it carries ownership and status.

Worked example

This is a synthetic teaching example, not SEOryon performance data.

A mini Academy has eight pages:

1. Academy hub;
2. SEO fundamentals;
3. search intent;
4. keyword research;
5. crawlability;
6. internal linking;
7. an internal-link checklist;
8. an older internal-links guide.

The crawl finds two orphans: the checklist and the old guide. Three anchors say learn more. The fundamentals lesson links to an old route that redirects to search intent. Both the old guide and Step 16 target the same task. The global footer links to all eight pages with keyword-heavy anchors.

Diagnose before changing

The checklist is valuable and belongs to Step 16, so Step 16 should link to it as a downloadable working asset. The old guide overlaps the lesson and has no distinct outcome, so valuable material should be merged into Step 16 and the old guide redirected. The source link should point directly to its final destination. The three vague anchors should describe their destination in context.

The footer does not need eight course links. It should link to the Academy hub. The hub and lesson navigation carry the sequence. Contextual links connect prerequisites and tools.

Repaired graph

• Home links to Academy.
• Academy lists all published lessons in order.
• Fundamentals links to search intent with a descriptive anchor.
• Search intent links to keyword research.
• Crawlability links to internal linking where discovery paths are explained.
• Internal linking links to the checklist and the next lesson.
• The old guide redirects to Step 16.
• Every source link is updated to its final canonical destination.

The page count decreased by one while the graph became more useful. That is a successful outcome.

Automation with safeguards

Automation can recommend links, detect unlinked mentions, and validate destinations. It should not publish every semantic match.

Require:

• a canonical destination registry;
• locale matching;
• tenant isolation;
• exclusions for account, legal, and sensitive states;
• per-page and per-template review limits;
• idempotency so reruns do not duplicate links;
• a dry-run diff;
• rollback data;
• approval for high-impact templates;
• monitoring after deployment.

For SEOryon, the safe unit is not "keyword found." It is "eligible source, approved destination, useful relationship, crawlable rendering, and verified ownership." If any element is missing, recommend rather than publish.

Monitoring the graph

Review the graph after migrations, navigation changes, bulk publishing, consolidation, and locale launches. Monitor:

• orphan count by indexable page type;
• broken and redirected internal links;
• final canonical mismatch;
• important-page click depth;
• links per template and per content section;
• newly competing destinations;
• links to noindex or unauthorized pages;
• user clicks on key next steps;
• stale owner or last-checked dates.

Use trends, not vanity totals. A drop in link count can be healthy after removing a sitewide footer grid. A rise can be healthy after connecting a new Academy. Interpret the change against the architecture decision.

Common mistakes

Linking every keyword mention

Repeated links interrupt reading and can send users away before the current explanation is complete.

Treating navigation as PageRank sculpture

Design for user choice and durable information architecture. No public formula lets you calculate internal authority precisely.

Keeping links to redirected URLs

Redirects preserve journeys during change. They are not the preferred permanent destination for links you control.

Fixing orphans with the sitemap only

A sitemap supports discovery but does not create a visible user path or explain relationships.

Linking to the wrong locale or tenant

Cross-locale links can confuse the experience. Cross-tenant links can expose data and become a security failure. Validate both before publication.

Using breadcrumbs that disagree with the page

The visible hierarchy, structured data, canonical owner, and navigation should tell the same story.

Exercise: audit one learning path

Choose one hub and eight destination pages. Export every link among them into the worksheet. Find:

1. one missing prerequisite path;
2. one vague anchor;
3. one redirect hop;
4. one destination with questionable ownership;
5. one unnecessary template link;
6. the click depth of each page from the hub;
7. one user action that should be measured.

Write a proposed change, owner, rollback, and validation check for each issue. Do not publish until each destination's status, canonical, locale, and access state are known.

Evaluation rubric

Final checklist

• Every link serves a stated user need.
• The destination is the canonical owner of that task.
• Links use an a element with an href where crawling is intended.
• Anchor text describes the destination naturally.
• Important pages have visible navigation or contextual paths.
• Prerequisites and next steps are connected.
• Breadcrumbs reflect a real user hierarchy.
• Orphan reports reconcile crawl, sitemap, analytics, and search data.
• Broken destinations are repaired or removed.
• Links point directly to final canonical URLs.
• No link crosses a tenant or authorization boundary.
• Locale links lead to equivalent available content.
• Template links are useful on every page using the template.
• Automated changes are idempotent and reviewed.
• A dry run and rollback exist for bulk changes.
• Link graph reports include an owner and last-checked date.
• User interaction is measured where it represents a meaningful next step.
• No arbitrary link count or anchor ratio is presented as a ranking rule.

Frequently asked questions

How many internal links should I add to a page?

There is no official ideal count. Add the links needed to complete the current task and offer useful next steps. Review unusually low or high counts as diagnostic signals, not automatic failures.

What is the best anchor text for internal links?

Use concise language that describes the destination in the context of the sentence. Avoid vague anchors and forced repetition. Natural variations are useful when they accurately reflect different relationships.

Are breadcrumb links enough?

No. Breadcrumbs express hierarchy. Contextual links can express prerequisites, evidence, comparisons, tools, and next actions that hierarchy alone cannot.

Does internal linking increase rankings?

Links support discovery, navigation, and understanding, but no individual link guarantees ranking. The destination still needs to be indexable, useful, accurate, and the right owner for the task.

Should I link to noindex pages?

Only when users genuinely need them. A noindex page can be part of a product journey, but it should not be presented as an organic landing-page owner.

How often should I audit internal links?

Audit after structural releases and on a cadence proportional to change. A frequently updated marketplace needs more frequent checks than a stable ten-page site. Triggered checks are usually more useful than a universal calendar rule.

Sources and methodology

Research combined official Google documentation, a disclosed vendor study, and recurring community questions about link counts, anchors, and orphan pages. Community discussions shaped the questions, not the factual claims. Sources were checked on 13 August 2026.

1. Google Search Central, Link best practices for Google. First-party guidance on crawlable link markup, anchors, context, and linked images. It does not disclose ranking weights.
2. Google Search Central, SEO Starter Guide. First-party guidance on site organization, navigation, and useful links. It does not establish an ideal link count.
3. Google Search Central, Breadcrumb structured data. First-party guidance on visible hierarchy and breadcrumb markup. Eligibility does not guarantee display.
4. Ahrefs, Link Rot Study. Vendor study reporting 66.5% link rot among links to 2,062,173 sampled websites from a January 2013 crawl over nine years. It is not a universal current rate or an internal-link-only study.

Previous: Entities and Topical Architecture
Next: E-E-A-T, Authorship, and Reputation