VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

How Web Links Fail | Why a Click Can Reach the Wrong Page, the Wrong Version or Nothing at All

Three secondary students review textbooks and notes together at a classroom table.

Web links fail in more ways than a familiar 404 page.

A link can technically open and still be wrong. It can land on a page with the same topic but the wrong canonical owner. It can reach an old version after a site migration. It can redirect through several intermediate URLs and finish somewhere the author never intended. It can use vague anchor text so the reader does not know what will happen before clicking. It can point to a page that still exists but no longer supports the claim the surrounding sentence makes.

That is why link quality is not a yes-or-no question about whether the browser displays a page.

A reliable link preserves a relationship between two pieces of information: the source page tells the reader why the destination matters, the URL or redirect path reaches the intended resource, and the final resource still contains the content or function the link promises.

This master guide owns the general failure architecture for ordinary web links. The three companion explainers go deeper into link target verification, redirect chains and link context and anchor text.

The article does not replace eduKateSG’s existing owners for web crawling, persistent identifiers, deep-link localisation or document cross-references. Those specialised articles keep their jobs. This cluster is about the everyday integrity of links as a reader-facing navigation and evidence mechanism.


The Core Idea: A Link Is a Promise About Destination

When a reader sees a link, the surrounding page makes an implicit promise.

It says: follow this path and you will reach something relevant to the words you just read.

A link therefore has at least three states to preserve.

  • Source intent — what the linking page says the destination will provide.
  • Routing — where the URL and any redirects actually send the reader.
  • Destination state — what the final resource currently contains and whether it is the correct version or owner.

Most link audits check only the middle state.

They ask whether the server returns an error.

That is useful, but incomplete.

A response can be successful at the HTTP level and still break the informational promise.


Failure Mode 1: The URL Returns 404

This is the visible failure everyone recognises.

The target resource is not found at that address.

A broken internal link creates friction for readers and can also interrupt crawling and site structure.

The repair depends on what happened: the page may have moved, been deleted, been merged into another owner or never existed at the typed path.


Failure Mode 2: The URL Returns a Page, but Not the Intended Page

This is harder to notice.

A link can return HTTP 200 and still land on the wrong content.

For example, a tuition article may link to a broad Mathematics hub when the sentence promises a specific algebra diagnostic. The destination exists; the semantic match is weak.

This is why target verification checks identity and content, not only reachability.


Failure Mode 3: A Site Migration Sends Everything to the Home Page

A common emergency repair is to redirect many old URLs to one surviving page.

The broken-link count may fall.

The reader experience can become worse.

A specific link about Secondary 1 Mathematics that now opens the home page has not preserved its original meaning.

A redirect should preserve resource identity closely enough that the final destination still satisfies the original link intent.


Failure Mode 4: The Link Reaches a Stale Version

Version drift can be invisible.

The old page still exists, looks professional and may even rank well.

But the site has a newer canonical owner with updated facts, corrected structure or a changed policy.

A working link can therefore compete with the site’s current architecture.

For eduKate’s one-intent-one-owner rule, this matters directly: internal links should reinforce the current canonical owner rather than keep obsolete parallel pages alive.


Failure Mode 5: The Destination Changed Meaning Without Changing URL

URLs are not always immutable identifiers.

A page can be rewritten in place.

If an external source originally supported one claim and later becomes a broader landing page, the link still resolves while the evidence relationship weakens.

This is especially important for references, policy pages, product documentation and fast-changing information.


Failure Mode 6: A Redirect Hides the Move

Redirects are useful because resources move.

HTTP defines several redirection status codes, and Google Search also treats permanent and temporary redirects differently when interpreting canonical destination.

The problem is not redirection itself.

The problem is forgetting that a redirect is part of the link path.

When a link has moved permanently, the author may eventually want to update the source page to point directly to the stable destination rather than depending forever on an unnecessary intermediate hop.


Failure Mode 7: Redirect Chains Become Long

A page moves from A to B, later from B to C, then from C to D.

The original source still links to A.

The browser may follow the chain successfully, but each hop creates another place where routing can fail, another piece of history to understand, and extra latency.

Long chains also make maintenance harder because the actual final destination becomes less obvious to authors.

Read How Redirect Chains Work for the full mechanism.


Failure Mode 8: Redirect Loops Trap the Reader

A points to B.

B points to C.

C points back to A.

No amount of successful redirect configuration helps if the path never reaches a usable resource.

Loops are a routing failure rather than a missing-content failure.


Failure Mode 9: Permanent and Temporary Moves Are Confused

HTTP 301 and 308 represent permanent moves.

HTTP 302 and 307 represent temporary redirects, although the method-preservation details differ.

Google Search documents that permanent redirects are used as a signal that the target should become canonical, while temporary redirects generally preserve the source URL as the canonical signal.

Using the wrong type can therefore communicate the wrong lifecycle intention.


Failure Mode 10: Redirects Change Request Behaviour

For ordinary reading links, this distinction may be invisible.

For forms, APIs and other requests, redirect status semantics can affect whether the original request method is preserved.

RFC 9110 distinguishes 307 and 308 as method-preserving redirects.

This is one reason technical teams should not treat every 3xx code as interchangeable.


Failure Mode 11: The Fragment Identifier No Longer Exists

A link may include a fragment such as #frequently-asked-questions or #sources.

The page itself still loads.

The reader lands at the top because the section identifier was renamed or removed.

The URL works at page level but fails at section level.

This is common after heading restructures and content migrations.


Failure Mode 12: Query Parameters Change the Destination State

A URL can contain parameters controlling language, filters, tracking, search or application state.

Copying the wrong parameter can send the reader to an empty result, a personalised view or an expired state.

When the parameter is not necessary for the promised destination, a cleaner stable URL is often easier to maintain.


Failure Mode 13: Authentication Changes What the Reader Sees

An author tests a link while logged in.

The public reader opens the same URL and receives a login page, permission error or different content.

The link may be valid for the author and useless for the intended audience.

Verification should include the audience state that matters.


Failure Mode 14: Geo, Locale or Device Routing Changes the Page

A destination may behave differently by country, language, device or application.

That can be intentional.

It becomes a link failure when the route changes the meaning the source page promised.

The existing eduKate translation estate owns deep-link and localisation details; this master simply preserves the principle that the final reader-facing destination should match intent.


Failure Mode 15: The Destination Exists but Is No Longer Indexable

An internal link can point to a page that is intentionally private, noindexed or otherwise outside the public search architecture.

That may be correct for a utility page.

It is wrong when the link is supposed to strengthen a public canonical content owner.

For SEO work, link validity includes the publication state relevant to the intended role.


Failure Mode 16: The Link Points to a Duplicate Owner

Two pages can answer almost the same question.

Internal linking may accidentally keep both alive by sending authority and readers to both.

This is not a classic broken link.

It is an architecture collision.

eduKate’s canonical-owner rule turns link review into a semantic task: the destination should be the chosen owner for that intent.


Failure Mode 17: The Anchor Text Is Too Vague

W3C accessibility guidance encourages link text that describes link purpose so users can decide whether to follow it.

‘Click here’, ‘read more’ and ‘this page’ may make sense inside one sentence, but they provide little information when links are scanned out of context.

A stronger link names the destination or the purpose.

That mechanism is developed in How Link Context Works.


Failure Mode 18: The Anchor Text Overpromises

Descriptive anchor text can still be wrong.

A link labelled ‘Complete Secondary 1 Mathematics Syllabus’ should not land on a short tuition advertisement.

The more specific the anchor, the stronger the obligation to match it.


Failure Mode 19: Identical Anchor Text Leads to Different Purposes

A page with six links all labelled ‘Learn more’ forces the reader to reconstruct context repeatedly.

For screen-reader users who navigate a list of links, the ambiguity becomes even more pronounced.

W3C notes that link purpose should be identifiable from link text or its programmatically determined context, and recommends descriptive naming where possible.


Failure Mode 20: Different Anchor Text Hides the Same Destination

This is not always wrong.

The same canonical page can serve several nearby contexts.

But a site can become confusing if different descriptions imply that the destination contains substantially different material.

Anchor language should be accurate enough that the shared destination still makes sense from every source.


Failure Mode 21: The Link Opens a New Context Without Warning

Opening a new window or launching an application can disorient users.

W3C lists advance warning about new windows as an accessibility consideration.

The behaviour should be deliberate and proportionate to the task.


Failure Mode 22: The Link Downloads a File Without Making That Clear

A link to a PDF, spreadsheet or other file can be perfectly valid and still surprise the reader.

Where file type, size or download behaviour matters, the surrounding text should make it clear.

The same principle applies to links that initiate mail, messaging or other application actions.


Failure Mode 23: The Destination Is Technically Reachable but Too Slow or Unstable

Link integrity includes practical reachability.

A repeatedly timing-out source may function during one audit and fail for readers later.

For important references, consider whether an authoritative alternative or persistent identifier exists.

Do not replace a primary source casually; preserve authority first.


Failure Mode 24: External Sources Disappear

Link rot is unavoidable over long time periods.

Organisations restructure sites.

Documents move.

Projects close.

A robust article therefore benefits from naming the source in visible text rather than making the URL carry all identity.

When a source moves, the author can often find the document again because the title and issuing body remain known.


Failure Mode 25: A Citation Link Is Reachable but No Longer Supports the Claim

This is more serious than cosmetic link rot.

The source may still exist but the cited fact is gone, qualified differently or superseded.

A claim-to-source relationship needs semantic revalidation when the evidence changes.

The Summary Traceability cluster provides the deeper provenance mechanism.


Failure Mode 26: Links Are Checked Once and Assumed Permanent

Websites are living systems.

A clean launch audit says something about that moment.

It is not a lifetime guarantee.

Review cadence should follow consequence and change rate.

Critical internal navigation and source links deserve more attention than low-impact peripheral links.


Failure Mode 27: Automated Checking Replaces Human Verification

Automation is excellent at finding status codes, redirect chains and large-scale patterns.

It is weaker at deciding whether the final page truly fulfils the semantic promise made by the source sentence.

The strongest workflow combines automated discovery with targeted human judgment.


Failure Mode 28: Human Review Ignores Scale

The opposite failure is manually opening a few favourite pages and assuming the site is healthy.

A large estate needs systematic discovery.

Automation finds candidates.

Humans resolve semantic questions.


Failure Mode 29: Repair Creates a New Collision

A broken link is found.

The editor replaces it with the first related page available.

That page turns out to be a secondary support article rather than the canonical owner.

The broken URL disappears while the architecture becomes less coherent.

Repair should route to the correct owner, not merely to any page that opens.


Failure Mode 30: Old Links Are Left in Place After Permanent Moves

Redirects are useful safety nets.

They should not become an excuse to leave known internal links permanently outdated.

When the final destination is stable and known, updating internal references reduces unnecessary hops and makes the site architecture clearer.


The Three Pillars of Reliable Linking

1. Target Verification

Check the final destination, page identity, semantic match and audience state. Read How Link Target Verification Works.

2. Redirect Integrity

Understand whether a page moved permanently or temporarily, where the chain ends and whether old references should be updated. Read How Redirect Chains Work.

3. Link Context

Make the purpose of the link understandable before the reader follows it. Read How Link Context Works.


A Practical eduKateSG Link Test

  • Source intent — what does the surrounding sentence promise?
  • Reachability — does the URL return a usable response?
  • Final destination — where does the link actually end after redirects?
  • Identity — is this the intended resource or merely a related page?
  • Canonical owner — is this the page that should own the intent now?
  • Content match — does the destination still contain the promised information?
  • Audience — can the intended reader reach it without the author’s login or special state?
  • Context — does the anchor text explain purpose accurately?
  • Behaviour — does the link open a page, file, application or new window as expected?
  • Freshness — is the destination still current enough for the claim?
  • Maintenance — should the source link be updated to a permanent final URL?
  • Evidence — when and how was the semantic match last checked?

Example: A Tuition Article

Suppose an article says, ‘Read the prerequisite map for Secondary Mathematics.’

A weak repair points the phrase to the general Mathematics homepage because that page is easy to find.

A strong repair routes to the actual prerequisite-architecture owner.

Both pages open.

Only one preserves the reader’s requested intent.


Example: A Research Reference

A sentence cites a government report about a current policy.

The original PDF moves and the old URL redirects to a publications landing page.

A status-code checker may mark the link healthy because the redirect resolves.

A human check asks whether the report itself remains reachable and whether the cited statement still appears in the current authoritative version.


Example: An Internal Hub

A hub contains thirty article links.

After several months of site rebuilding, five old URLs redirect, three pages have new canonical owners and one article has been retired.

The correct repair is not simply to preserve reachability.

The hub should be updated so every link represents the current knowledge architecture.


Example: Accessibility

A paragraph ends with ‘click here’.

The reader using a screen reader opens a list of page links and hears ‘click here, click here, click here’.

The destinations become difficult to distinguish.

Changing the anchors to meaningful phrases improves accessibility and also improves editorial clarity for every reader.


Example: Student Research

A student copies a URL into notes without writing the source title.

Weeks later, the link is dead.

If the note also records the author, document title and what claim it supported, the student has a much better chance of finding the source again or recognising an updated version.

Link literacy is therefore part of research literacy.


Across the eduKate Ecosystem

This general link-integrity cluster should strengthen, not replace, nearby owners. How Web Crawling Works owns discovery and crawling. How Persistent Identifiers Work owns stable identity when locations change. How English Works | The Cross-Reference owns the document concept. Deep Links, Universal Links and App Links owns localisation-specific routing. How Summary Traceability Works owns claim provenance and source paths.


Sources and Further Reading

RFC 9110 — HTTP Semantics

Google Search Central — Redirects and Google Search

W3C WAI — Understanding Link Purpose in Context

W3C WAI — Providing Link Text That Describes the Purpose of a Link


The Deeper Principle: A Working Link Preserves Meaning, Not Merely Transport

The web is built from paths between resources.

Those paths are useful only when the reader can understand why to follow them and arrives at the right destination.

A good link audit therefore checks more than whether a URL responds.

It checks whether the relationship still makes sense.

Reach the right page.

Preserve the right owner.

Describe the destination clearly.

And when the web changes, repair the path before the old link quietly becomes misinformation.


Frequently Asked Questions

What is the difference between a broken link and a wrong link?

A broken link cannot reach a usable target at the expected address. A wrong link may open successfully but reach a resource that does not fulfil the source page’s promise. Wrong links are often harder to detect because mechanical audits can mark them healthy.

Should every moved page have a redirect?

Only when there is a meaningful successor or a legitimate temporary destination. Redirecting a retired specific page to an unrelated home page can mislead readers. Preserve continuity where continuity exists; use an honest not-found state where no suitable replacement exists.

Can a redirect itself be considered a broken link?

Not automatically. Redirects are normal web infrastructure. A redirect becomes a problem when it loops, chains unnecessarily, communicates the wrong lifecycle, changes important request behaviour, or lands on a semantically wrong destination.

Why update internal links if redirects already work?

Current internal links should express current architecture. Pointing directly to stable canonical owners reduces unnecessary hops and makes the site easier to understand and maintain. Redirects can remain as compatibility paths for old external references and bookmarks.

What makes an internal link strategically wrong?

It routes an intent to the wrong owner. Examples include linking a general keyword to a duplicate page, sending a learning question to a commercial page when a canonical explainer exists, or continuing to reinforce an article that has been superseded.

How should external evidence links be maintained?

Keep the visible source identity—title, organisation and date where relevant—so the resource can be rediscovered. Recheck high-consequence claims when the destination changes or the source is time-sensitive. Do not assume that a surviving URL still supports the same proposition.

Are relative URLs better than absolute URLs for internal links?

Both can be valid depending on platform and architecture. Relative links can make some migrations easier; absolute links make the target explicit. The editorial standard remains the same: the resolved destination must be the correct current owner and must work for the intended reader.

What is a soft 404?

It is a page that behaves like missing content while returning a successful-looking response, often 200. A mechanical checker may see a healthy status even though the page tells the user nothing was found. Important link audits should inspect page content or other signals, not only status.

Do search engines and readers care about the same link failures?

There is overlap, but not complete identity. Readers care about clarity, relevance and successful navigation. Search systems also interpret redirects, canonicals and internal architecture. The strongest repair serves both by routing humans and crawlers to the correct current resource.

How can a link become misinformation?

When the surrounding sentence continues to make a claim that the destination no longer supports. The URL may still work. The source may still be authoritative. But if the relevant statement was removed, narrowed or superseded, the citation relationship has failed and should be repaired.

Should a large site check every link manually?

No. Use automation to discover mechanical failures and routing anomalies at scale, then focus human review on important internal owners, changed destinations, factual citations and ambiguous redirects. The audit method should concentrate judgment where semantics matter.

What should a repaired link record contain?

At minimum, keep the source page, old target, new target, reason for change and verification date for important repairs. For canonical-owner changes, record which intent moved and why. This creates continuity when the site is reviewed again later.


A Link Maintenance Rhythm

Treat link integrity as a living maintenance process rather than a one-time launch checklist.

  • After publishing — verify high-value internal and source links in the public reader state.
  • After migration or consolidation — audit redirects, canonical ownership and section-level links.
  • On a recurring basis — scan mechanically for failures, new chains and unstable external destinations.
  • After a factual or policy update — recheck the sources that support changed claims.
  • After finding one systemic failure — search for other links created by the same template, workflow or old owner.

This rhythm keeps the site from accumulating invisible routing debt. It also prevents audits from becoming enormous periodic rescue operations.

Continue the Series

How Link Target Verification Works | Check Destination, Identity and Content Before Trusting the Link

How Redirect Chains Work | Move Pages Without Hiding the Final Destination

How Link Context Works | Write Anchor Text That Explains Where the Reader Is Going

eduKateSG


What a Good Link Audit Should Not Claim

A clean crawl does not prove that every link is meaningful. A semantic sample does not prove that every page was checked. A successful redirect does not prove canonical ownership is correct. A descriptive anchor does not prove the source behind it is current. Keep these layers separate in reporting.

For a large estate, the strongest audit statement names its coverage: which content types were scanned, which mechanical conditions were tested, which high-value routes received human semantic verification and which unresolved classes remain. This makes the result repeatable and prevents a future editor from assuming a broader guarantee than the evidence supports.

The discipline is the same one used throughout this How X Works series: define the state, test the relevant mechanism, preserve the exception and avoid turning a narrow success into a universal claim.