Broken Links Test Page

How to test these links and compare results

Run your link checker on this page, or use individual examples in your browser automation tests. Wait at least three seconds before scanning so the dynamically added link is included.

Each section explains the example links and includes a results table. What the server does describes the response or failure associated with each example. Predicted in LinkScan describes its predicted classification, not a verified result. Other tools may classify the same link differently.

Some links deliberately return errors, trigger certificate warnings or fail to connect. External websites may change their behaviour over time.

For automated comparisons, the example URLs, server behaviour and LinkScan predictions are also available as JSON at /example/expected.json.

1. Protocol & Scheme Variations

Compare HTTPS, HTTP and special links that open apps or files instead of web pages.

• HTTPS link — The normal case: a secure, complete address. Nothing to fix.
• HTTP link — Insecure. Browsers upgrade it or warn the visitor. Use https://.
• WWW HTTPS link — A different host than the bare domain, so it costs an extra redirect. Link to the canonical host.
• Protocol-relative — Inherits the page's scheme. A legacy pattern; write https:// explicitly.
• Scheme-less link — Missing https://, so the browser treats it as a path under the current page and it breaks. Add the scheme.
• FTP Link — Browsers no longer open FTP. Serve the file over HTTPS.
• File system link — Points at the visitor's own disk, so it never works for anyone else. Remove it.
• Data URI — Inline data that is invalid here, and browsers block opening data: links. Link to a real file.
• Custom scheme (SMS) — Opens an app, not a page, so the scanner skips it. Make sure the number is right.

2. Absolute vs. Relative Paths

See how full URLs and relative paths point to different destinations.

• Absolute link — A full address that resolves the same from any page.
• Root-relative — Resolves from the site root. Safe when pages move.
• Path-relative — Depends on where the page sits, so it breaks when the page moves. Prefer root-relative.
• Dot-relative — Resolves inside the current folder (here /example/contact/), which does not exist. Prefer root-relative.

3. Internal vs. External Links

Test links within this site, across subdomains and to other websites.

• Internal link — A page on this site, under your control.
• Cross-subdomain — A separate host that may not exist or may be down. Check it like an external link.
• External site — Outside your control and can disappear any time. Re-check external links regularly.

4. Anchor Fragments

Compare working and missing jump targets on this page and another page.

• Working anchor — Jumps to an element with that id on this page.
• Broken anchor — No element has this id, so the click does nothing. Fix the id or remove the link.
• Bare hash — Jumps to the top of the page; usually a placeholder or a script hook. Use a button for actions.
• Absolute self anchor — Same page written with its full path. Works, but a plain #anchors is simpler.
• Anchor to a name attribute — Targets a legacy name attribute. Use an id instead.
• Other page anchor — The page is checked; the fragment on it is not.
• Other page, missing anchor — The page loads, but the jump target is missing, so the visitor lands at the top.

Dead same-page targets are checked in the DOM without a request. A fragment on another page is checked as that page without the fragment, so the last two links share one result.

5. Cyclic & Self Links

Try a link that reloads this page and an anchor that jumps to the top.

• Self-link — Points to the page you are on and just reloads it. Remove it, or mark it with aria-current="page".
• Top anchor — Browsers jump to the top of the page for #top even when no element has id="top".

6. Mail & Telephone

Compare complete and incomplete email addresses and telephone numbers. These links open apps rather than web pages.

• Valid mail — Opens the mail app with an address and subject. The scanner skips it.
• Invalid mail — No address, so it opens an empty draft. Add the address.
• Valid tel — International format that dials correctly on phones.
• Malformed tel — Letters can't be dialed. Use digits with a country code, like +14165551234.

7. Image & Media Links

A visible image can link to a working page or a missing one. Test the destination, not just whether the image loads.

• — Image link that works. Its name comes from the image's alt text.
• — Broken link in image. The image shows fine, but the target returns 404, which is easy to miss.

8. JavaScript-driven Links

This example uses JavaScript instead of a page URL. Check how your tool handles an action without a destination.

• JS Link — Not a real destination: fails without JavaScript and with middle-click. Use a button for actions.

9. HTTP Status Code Playground

Test successful responses, missing pages, access errors and slow replies. Some examples respond differently to HEAD and GET requests.

• 200 OK — The page loads. Nothing to fix.
• Working Page — Another healthy page.
• 404 Error — The page doesn't exist. Fix the address or add a redirect.
• 401 Unauthorized — Needs a login. Fine for members-only pages, broken for the public.
• 403 Forbidden — Access refused. Often a bot block or wrong permissions; open it in a browser to confirm.
• 410 Gone — Removed on purpose. Remove the link.
• 429 Too Many Requests — Rate-limited. Not necessarily broken; check again later.
• 500 Internal Server Error — The server failed. Check again; report it to the site owner if it persists.
• 502 Bad Gateway — A proxy got no valid answer from the server behind it. Usually temporary.
• 503 Service Unavailable — The server is down or in maintenance. Check again later.
• HEAD answers 405, GET answers 200 — The server refuses quick HEAD checks; the scanner retries with GET. Not broken.
• HEAD answers 404, GET answers 200 — The server answers HEAD checks wrongly, but the page works.
• Slow response (15 seconds) — Visitors may give up before it loads. Look at server performance.

Compare the recognised bot challenge with the plain 403 above, the login redirect with a direct login link, and the six-second success with the fast control and timeout. Live bot-header access and selected-request timing still need verification.

• 504 Gateway Timeout — Unlike 503, a 504 is counted as broken.
• HEAD blocked by a bot rule — HEAD 403 is retried with GET.
• HEAD not implemented — HEAD 501 is retried with GET.
• Works when clicked, blocked for scanners — FR56: requires the extension to read cf-mitigated: challenge. Confirm header access in the selected build.
• Rate limit with Retry-After — 429 with Retry-After: 30, every method
• Down for maintenance — 503 with Retry-After: 120, every method
• Login-gated page — FR56 recognises the login path reached from a different path.
• Slow but under the timeout — FR57: successful responses at or above 3000 ms are advisory; timeout stays Unknown.
• Direct login destination — An intentionally linked login path is not a login-redirect finding. Compare R-login-gate and the plain S9-403.
• Fast successful response — Negative slow-response control. Network conditions can make even this route slow. Timing covers the selected request to headers, not page rendering.

10. Accessibility Attributes

Test links with missing destinations, empty names and vague text. Compare what is visible on the page with the name available to screen readers.

• Proper ARIA — The label names the destination. Good.
• Empty link — No text and no label, so screen readers just say "link". Add text or an aria-label.
• No href — Not a link at all and not keyboard-focusable. Add an href, or use a button.
• Empty href — Reloads the current page. Point it at a real destination.
• Whitespace href — Same as an empty href.
• Click here — Means nothing out of context. Describe the destination instead.
• READ more — Generic text. Say what there is to read.
• here — The label helps screen readers, but sighted visitors still see "here". Make the visible text descriptive.
• Label-only link — Has a name for screen readers, so LinkScan reports it as OK, but there is nothing visible to click. Add visible text.
• Privacy policy details — Named by other text on the page through aria-labelledby, so LinkScan reports it as OK. Valid, but the link itself has nothing visible to click.
• Image with empty alt — The link has no name at all. Give the image alt text that names the destination.
• Click here to send the developer a message — LinkScan flags only link text that is exactly a vague phrase, so this reports as OK. The text works, but it starts with filler. Lead with the destination.

11. Rel & Target Attributes

Compare new-tab links with and without rel attributes. A working destination does not tell you whether a tool checks these attributes.

• Correct _blank — Opens a new tab safely.
• Missing rel — In older browsers the new tab can control this page. Add rel="noopener".

12. Security & Mixed Content

Compare an HTTP link on this HTTPS page with a destination that has an untrusted certificate. Watch for HTTPS upgrades and browser security warnings.

• HTTP on HTTPS page — Sends the visitor from a secure page to an insecure one. Use https://.
• Bad SSL certificate — Self-signed, so browsers show a full-page warning. The site needs a trusted certificate.

13. Redirects

A redirect sends a request to another URL. A redirect chain passes through several URLs. A redirect loop repeats destinations without reaching a page.

• 301 Moved Permanently (one hop) — Works. Update the link to the new address to skip the hop.
• 308 Permanent Redirect (one hop) — Like 301. Update the link to the final address.
• 302 Found (one hop) — A temporary move. Fine to keep while the move is temporary.
• Redirect chain (two hops) — Each hop adds delay. Link to the final address.
• Redirect chain (three hops) — A longer chain, same fix: link to the final address.
• Redirect chain (twelve hops) — A browser may follow the full chain, while a checker may stop at its redirect limit. This is a long chain, not a repeating loop.
• Redirect loop — Pages redirect to each other forever; browsers show an error. Fix the redirect rules.
• Redirect to a missing page — Looks like a redirect, but it ends in a 404. Fix the redirect target.
• HTTP and www redirect — Two upgrades for one link. Use the canonical https:// address.

A redirect to a recognised not-found path returning 200 is predicted Broken; a deep-path redirect home is advisory. Direct 200 error bodies are not read. The ordinary one-hop links above are negative controls.

• 307 Temporary Redirect (one hop) — 307 → /privacy-policy/
• Soft 404 (error page with status 200) — The body is not read. A direct 200 with error text remains a missed problem in v2.1.
• Soft 404 by redirect to the home page — FR56 detects deep-path redirects ending at the site root, even with one hop.
• Redirect to a 200 not-found URL — The recognised final path supplies the signal, not its response body.
• Direct 200 not-found URL — Negative redirect control: a direct 200 error page remains undetected even at a recognised path.

14. Network Failures

Test failures that happen before a page returns an HTTP status: missing domains, invalid certificates and connections that do not respond.

• Domain that does not exist — The reserved .invalid suffix is classified as a local address without a request from this public page, not as a DNS failure.
• Expired SSL certificate — Browsers block the page until the site renews its certificate.
• Certificate for another host — The certificate names a different domain, so browsers block the page.
• Port that does not answer — The connection times out. Check the port in the address.

The local/private examples below are for scanning, not for browsing your machine or network. The public HTTPS link in §1 is a negative control. Same-site exemptions apply only when the scanned page is itself local, private, staging or preview—not to every link on that page.

• Localhost address — Do not click unless you intend to access your own machine. LinkScan blocks this linked request on a public page.
• Private IPv4 address — Do not click unless you intend to access your private network. Compare the public HTTPS control in §1.

15. Duplicates & Hidden Links

The same broken address appears three times: once visibly, once inside a closed section and once hidden with CSS. Check whether your tool finds all three locations and combines duplicate URLs.

• Repeated link (visible) — One broken address used in several places. Fixing it once fixes every copy.
• Hidden until a visitor opens the section, so it is easy to forget:
  Closed section

  Repeated link (inside a closed section)
• Repeated link (display: none) A copy hidden with CSS: invisible to visitors, but still in the page. Remove stale hidden links.

16. Edge-case URLs

Compare URLs with tracking parameters, different casing, encoded spaces and international characters. Some resolve to the same page; others may not.

• Tracking parameters — A separate address from the clean one. Keep tracking off internal links.
• Uppercase path — Paths are case-sensitive on many servers, where this would be a 404. This site's host redirects it to /contact/ instead, which costs a hop. Match the real path.
• No trailing slash — Often costs a redirect. Match the site's canonical form.
• Encoded space in query — Spaces must stay encoded as %20.
• Read the complete guide to finding, fixing and preventing broken links on every page of your website, including redirects, missing anchors and slow external resources — Very long link text. Works, but it is hard to scan and gets cut off in result lists. Keep link text short and specific.
• Internationalized domain — A non-ASCII domain; browsers convert it to punycode (xn--…).
• Added by a script two seconds after the page loads. Wait for the page to finish before you scan:

Links inside graphics, web components and frames

These examples place links inside SVGs, image maps, web components and embedded frames. Some tools may not detect them. Each points to a missing page; LinkScan's predicted coverage is listed in the table below.

• A link inside an SVG graphic:
• A clickable area of an image map:
• A link inside a web component's shadow DOM:
• A link inside an embedded frame:

Preview links and controls

These same-site query examples exercise the implemented preview rule without relying on a third-party staging host. Specific preview patterns and same-site staging labels are recognised, not every draft URL. Broken and Unknown outcomes take precedence. Slow advice measures the selected request to headers, not page load.

• Same-site preview query — The exact preview=true query triggers advice on this public page.
• Non-preview query control — preview=false does not match the implemented preview=true rule.
• Preview URL with an HTTP error — Broken HTTP outcomes take precedence over preview advice.
• Preview URL temporarily unavailable — Unknown HTTP outcomes also take precedence over preview advice.