EPUB Link Checker

The EPUB Link Checker is built for the moment after you have produced an EPUB and need to know whether its navigation holds together. It reads the file in your browser, walks every page, the navigation document and the NCX, and reports which links, anchors and images point at something that really exists inside the book. Nothing is uploaded; the EPUB is unpacked and inspected locally, and the results disappear when you close the tab.

It suits authors and publishers preparing a store submission, where a footnote that jumps nowhere is a defect readers notice, and conversions from Word, InDesign or Markdown, where relative paths and cross-references quietly break.

The same scan helps when you inherit an EPUB from someone else, or pull an older title out of an archive and want to confirm its links and embedded images still resolve before you reuse or redistribute it.

The flow is one step: drop or select a single .epub file. The tool unzips it, reads the package, and opens each page listed in the spine, then the navigation document and the NCX. From each it collects link targets (href on <a> and <area>), image sources (src on <img>, and href or xlink:href on SVG <image>), and table-of-contents targets. Results appear grouped into broken, external and valid, and the full list can be exported as a plain-text report.

Each internal link is resolved relative to the file it sits in, so ../Images/fig.png from a page in Text/ points at Images/fig.png. A link is broken when the target file is not in the book, or when it names an anchor, such as chapter2.xhtml#section4 or #note3, and the target page has no element with that id. A target that exists in the ZIP but is missing from the manifest is flagged as a warning, since some readers only load listed files. Links that start with a scheme such as http://, https:// or mailto:, and protocol-relative // links, are listed as external and not visited, and DRM-protected files are refused.

What it does not do: it does not follow JavaScript-driven navigation, does not fetch or validate external URLs, and does not check references inside stylesheets, such as fonts or background images.

A valid result means the target file and anchor exist, not that the link lands where the author meant. A footnote link that points at the wrong note, or a table-of-contents entry with the wrong label, passes as long as its target exists. Spot-check a few links in a real reader.

Anchor ids are matched exactly, including case, so #Note3 and #note3 are different targets; ids are looked up in id attributes and in the legacy name attribute on <a>. A long book can produce a large list: the screen shows the first 50 problems, and Show all or the downloaded report gives the complete picture.

External links shown as unverified are neither failures nor guarantees; the tool cannot tell a live URL from a dead one, so check them by hand. Only check files you own or have permission to inspect.

Start with the broken links: open each source file in your editor, confirm the target path or anchor is really wrong, then correct the relative path, restore the missing file, or add the missing id. Next, add any files flagged as missing from the manifest to the package. Then work through the external list by hand, opening the URLs to confirm they still resolve.

Re-run the scan after edits and aim for zero broken links, then preview the book in an actual e-reader to confirm the navigation feels right rather than merely passing a structural check. The text report is a useful record to keep alongside the file as you iterate.

For a regulated or reference publication, check the affected links in your reading software too.