Skip to main content
Back to Blog
·Roberto·12 min read

EPUB Accessibility Checklist: Test Your Ebook Before Publishing

EPUB Accessibility ChecklistEPUB QASelf-PublishingEbook Accessibility

An EPUB can open without errors and still be unusable for someone navigating by headings, listening with a screen reader, or enlarging the type. An export button proves that a file was created; it does not prove the book can be read. This EPUB accessibility checklist gives independent nonfiction authors a practical release test they can run before uploading a book to a store or delivering it to a customer.

You will get a twelve-part inspection workflow, a downloadable QA worksheet, and a complete fictional example showing how to find and resolve problems. The checklist is not a certification process or a claim that every ebook made with a particular tool meets an accessibility standard. It is designed to expose problems while you can still correct the manuscript.

Download the free EPUB accessibility checklist

Download the editable EPUB accessibility QA worksheet (CSV). The sheet contains twelve columns, six clearly fictional example findings, and eight blank rows for your own checks. Keep a separate row per issue and a retest date when it is fixed. This is a manual audit log, not a scanner or automated certificate.

Want to draft a source-grounded nonfiction book first? Scribora can build an editable outline and chapter drafts from your permitted notes, documents and transcripts, then export PDF or EPUB. Start with a free outline and sample chapters; you still need to inspect the final exported files and perform the checks below.

What an EPUB validator can and cannot tell you

Separate file validity, automated accessibility checks, and human reading experience. These are different tests. Theofficial EPUBCheck project checks EPUB format conformance. It is the right first step for malformed packages, documents and metadata, but a passing EPUBCheck report is not proof that your heading structure, image descriptions, or reading order make sense to a person.

Ace by DAISYperforms automated accessibility checks and gives you findings to investigate. Its maintainers explicitly say it cannot make a complete accessibility judgment, issue a pass/fail certification, or repair the content for you. The DAISY SMART manual-reporting tool can guide human checks alongside Ace. In other words: run tools, then read and navigate the actual book. Do not advertise a percentage score as proof.

The normative reference is the W3C EPUB Accessibility 1.1 Recommendation, published October 17, 2024. W3C also published an EPUB Accessibility 1.2 Candidate Recommendation Draft on November 11, 2026. That later document is a draft, not a new Recommendation; do not silently treat it as finalized certification criteria. This checklist is an author workflow, not a substitute for assessing full W3C conformance.

The twelve EPUB accessibility checks to run

  1. Preserve a clean source. Save the editable manuscript and the exported EPUB you are testing. Record the exact version and date. If you fix an issue in a different file, your result no longer describes the published artifact.
  2. Validate the package. Run EPUBCheck on that exact EPUB. Keep the report, fix errors, and rerun. Do not accept a reader opening the file as a format-validation result.
  3. Inspect the table of contents. Open the navigation menu and jump to every major chapter and appendix. Confirm that labels, targets and order agree with the book, including its final sections.
  4. Check heading structure. Browse by headings with assistive technology, or inspect the heading markup. A visually large bold line is not necessarily a heading; headings should describe a sensible hierarchy.
  5. Test reading order. Listen through the first chapter, a chapter transition and every unusual layout. Verify that callouts, footnotes and sidebars are encountered where their meaning belongs.
  6. Review images by purpose. An informative image needs an equivalent description; a purely decorative image may need an empty text alternative. Do not write generic descriptions such as “image” or repeat text already adjacent to a decorative illustration.
  7. Expand complex diagrams. A trend chart often needs its takeaway, axes and key values expressed in nearby text or a data table. A ten-word alt attribute cannot necessarily replace a complex figure.
  8. Inspect actual tables. Tables should use marked-up row or column headers. If a table becomes confusing when read cell by cell, consider splitting it or rewriting it as a short list.
  9. Check links and notes. Test all links and backlink navigation for references or endnotes. Link text should explain the destination rather than saying “click here” repeatedly.
  10. Resize and reflow. Change the reader's font size and spacing. Watch for clipped text, unexpected horizontal scrolling, missing captions, and content whose meaning depends on exact page position.
  11. Verify language and meaningful metadata. Inspect the book language, title and relevant accessibility discoverability information. Do not declare features such as alternative descriptions or accessibility conformance without actually checking them.
  12. Retest after corrections. Export again, rerun EPUBCheck and Ace where applicable, and repeat targeted manual checks on the final file. Record what passed, what remains blocked, and why.

The W3C image alternatives tutorial explains the difference between decorative, informative and complex images. Its accessible tables tutorial explains why header cells and their relationships matter to screen-reader navigation. These are useful authoring references, not ebook-specific validation results.

An EPUB quality check has three layers: valid file, automated accessibility findings, and human navigation and reading tests. Illustration, not a certification badge.

Worked example: fixing a fictional EPUB before launch

Imagine Lena, an entirely fictional workshop organizer, has written a guide called Run Your First Community Repair Session. Its manuscript contains six chapters, a venue checklist, two data tables and a volunteer flowchart. Nothing here is a real customer story, an observed product test or a claimed Scribora export defect. The examples demonstrate how a release log makes decisions reviewable.

Lena first exports a draft EPUB. EPUBCheck reports no format errors in this hypothetical scenario. A visual inspection looks fine too. Yet the navigation menu exposes only three of six chapters. She records the issue as a release blocker: a reader cannot reliably find the missing sections. She corrects the heading and navigation structure in her source, exports a new file, and verifies all six destinations rather than just the first fixed link.

The volunteer flowchart is the second issue. The image description merely says “volunteer process.” That name does not convey the decision branches. Lena adds a short equivalent explanation immediately after the image: “If the participant needs tools, direct them to equipment check-in; otherwise send them to the intake desk. Both paths reach the session lead before work begins.” The text is a fictional example of an equivalent, not safety guidance for real repairs. The diagram's description and surrounding prose must be checked against the actual image, not copied mechanically.

The third issue is a two-column table of roles with cells that make sense visually but lack proper column headers. Lena reviews the markup and restores the role and responsibility headings. For a narrow-screen device she also tests whether the information remains usable when text is enlarged. She could simplify the table into a sequence of role definitions if the layout still breaks.

Fictional issue log excerpt from the free worksheet
CheckInitial findingDisposition
TOC and navigationThree missing chapter linksBlock release; repair source; retest six links
Flowchart alternativeNon-informative descriptionAdd equivalent decision paths in prose
Role tableMissing marked-up headersRestore headers and test enlarged text

This illustrates why a tool-only “green” result is insufficient. A machine can flag some missing attributes, but it cannot determine whether an actual description says what the diagram means. Nor can it decide that the book's chapter titles are a useful journey for the intended reader. Those require editorial judgment.

How to use the worksheet without inventing audit results

Each row records the inspected file version, checkpoint, location, method, observed result, severity, correction, owner, and retest. Start with the example rows, but replace every fictional finding with an observation made on your actual file. Not tested means not tested; it does not mean pass. Mark a chapter-level blocker as unresolved until you have tested the rebuilt EPUB.

Use three distinct statuses: pass when you actually verified the result, fail when you observed a defect, and not testedwhen you have not run the check. A fourth not applicable status needs a reason; “not applicable” should not become a convenient way to hide diagrams or tables you forgot to inspect. The sample CSV has invented data only. It does not certify anything.

An example release decision could read: “Draft 0.2 is blocked: navigation issue A1 remains open, while image description B2 is ready for retest.” After repairing the source, Lena would record a new file version and date for every retest. She would not claim the book passed EPUB Accessibility 1.1 simply because the twelve worksheet cells were filled in.

Where Scribora fits into the EPUB workflow

Scribora helps turn permitted material into an original nonfiction manuscript: source analysis, an editable outline, chapter review, and PDF or EPUB export. It is not a replacement for EPUBCheck, Ace, reader testing, full accessibility conformance assessment, or professional remediation. If you already have a complete and accessible manuscript and just need validation, start with the independent tools above.

If the obstacle is that your notes, handouts or research documents do not yet form a coherent book, use the ebook outline template to identify the chapter outcomes and try Scribora's free sample workflow. Review the source-based drafts for correctness, then apply this EPUB accessibility checklistto your own final exported file. The promise is an inspectable workflow, not guaranteed certification.

Publish only the file you actually tested

Before placing the download on a sales page, store or partner site, record the final file version, validator reports, known limitations, and who signed off each manual test. Check the delivered file again after a platform converts or processes it; retailer pipelines can alter what your customer receives. If you are still choosing a destination, consult the ebook distribution guide. Do not mistake a store accepting an upload for proof of accessibility.

The most useful release question is not “does my EPUB look polished?” but “can a reader reach, understand and use the information when their reading method differs from mine?” Run the machine checks, inspect meaningful content manually, repair the source, and publish the specific build you retested.

Sources verified November 16, 2026: W3C EPUB Accessibility 1.1 Recommendation (2024), W3C 1.2 Candidate Recommendation Draft (November 11, 2026), EPUBCheck, DAISY Ace and SMART, and W3C WAI image and table tutorials. The example book, author, findings and outcomes are invented for instruction; no production audit or user result is represented here.

Publishing on Amazon? The KDP ebook formatting preflight adds Kindle navigation, device preview and store-upload checks to this broader accessibility audit.

Ready to create your own ebook?

Turn your own notes, transcripts, links, documents, and audio into an outline and editable chapter drafts.

Start for Free

Not ready yet? Get the free 30-page guide — how to ship a non-fiction ebook without the AI slop.

Pricing for Kindle? Work out your KDP royalty — free, no signup, all nine Amazon stores.