Interactive eBook BuilderInteractive eBook Builder
$59 · unlimited access, unlimited eBooksGet it for $59

· Updated · By · 9 min read

EPUB Accessibility: Automated vs. Manual Checks

epubaccessibilitytesting

I use automated checks to find EPUB errors and manual review to check whether readers can use the book. Start with EPUBCheck and Ace by DAISY, then review 3 areas: image descriptions, reading order, and navigation. A passing tool report is not proof of accessibility conformance.

Quick Comparison

Check Automated tools Manual review
Images Flag missing text alternatives. Check whether descriptions explain the image’s purpose.
Structure and reading order Find markup errors and possible sequence problems. Check whether the content follows a logical flow.
Navigation Detect missing or broken references. Check labels, destinations, and ease of movement.
Evidence Record machine-tested results. Document judgments and behavior in tested reading systems.

My release process is simple: define the conformance target, test the exported EPUB, fix issues, and repeat the affected checks after changes. Keep a log of versions, findings, and anything left untested. Make sure accessibility metadata and claims match those records.

AI can help polish draft content, but it does not replace EPUB testing. If you use Interactive eBook Builder for that preparation, Notion AI is required separately; the workflow does not imply built-in EPUB export.

EPUB testing

What automated EPUB checks detect

EPUBCheck verifies structure; Ace checks machine-testable accessibility.EPUBCheck checks the EPUB’s technical structure, including package metadata, malformed markup, and broken references. Ace by DAISY checks accessibility requirements that software can test. You’ll need to review the findings that these tools can’t resolve.

Automated tools can inspect language declarations, heading markup, image alternatives, reading order cues, and navigation elements. They can check what’s present, but they can’t fully judge whether it’s correct.

Automated findings that need human judgment

Treat each report as a review queue, not a final verdict. Human review checks whether the EPUB works for readers.

Issue Automated detection Human judgment required Typical remediation
Image alternatives Flag missing alt text. Decide whether the image needs a meaningful description. Add or revise the description.
Heading hierarchy Flag possible heading hierarchy problems. Check whether headings match the content’s organization. Adjust heading levels and wording to match the structure.
Reading order Flag possible sequence problems. Verify that the content reads in the intended order. Reorder the content or adjust its structure.
Navigation labels and destinations Detect broken navigation references. Confirm that labels are clear and links point to the right place. Clarify labels and correct links to the intended content.

What manual EPUB review checks

After automated checks, manual review checks whether readers can use the EPUB.

Test the EPUB in supported reading systems and with assistive technologies. Check reflow where it applies. Record the EPUB build, operating system, reading-system and assistive-technology versions, and test steps.

Representative testing finds problems; it does not establish full conformance. Note what remains unevaluated.

Image descriptions that explain the content

Read each description in context. Check that it explains the image’s purpose and gives readers the information they need to understand the passage or complete the task.

Reading order that preserves meaning

Listen to the EPUB without relying on the page layout. Check that the reading order follows the intended flow and that headings match the structure. Even valid markup can produce a confusing spoken sequence if the source structure leaves the intended order unclear.

Navigation that helps readers move through the book

Check that table-of-contents entries have clear labels and lead to the correct sections.

Automated checks vs. manual review: key differences

The earlier sections listed the checks. Here’s how to separate issues a machine can detect from judgments about the reading experience.

Use automation for repeatable checks and human review for meaning and reader experience. Each method has limits, so keep your findings tied to what you tested.

Evaluation area Automated checks Manual review
Purpose Test rules a machine can evaluate. Assess meaning and reader experience.
Speed Run quickly across the EPUB. Take focused time and attention.
Limits Miss issues outside the rules they cover. Can miss content and behavior outside the test scope.
Content checks Flag structural issues. Judge whether the result works for readers.
Reader behavior in test environments Cannot fully predict the reading experience. Check behavior in selected environments.
Remediation Identify rule failures and possible problems. Weigh context and impact to choose the right fix.
Evidence Reports document what was tested. Review findings address requirements that need human judgment.

Use this comparison to decide which findings need fixes and which need manual verification. After each fix, rerun the relevant automated check and repeat the affected manual test.

Why passing tool checks does not prove conformance

A passing automated report means the EPUB met the tool’s rules. It does not mean the file meets every accessibility requirement.

A file can pass while an image description still misses the point, the reading order still confuses listeners, or navigation still slows movement through the book; human review must confirm meaning and usability.

Evidence for conformance claims

A tool report is evidence, not proof. Any conformance claim should identify the specific EPUB version reviewed, the tools used, and the tests run.

Use the report to guide manual review - not as the final verdict.

How to combine automated and manual EPUB checks

Automate first, then check the reading experience manually. Before testing, define your EPUB version and conformance target. Run validation and accessibility checks on the exported EPUB - not the source file. Sort findings into confirmed defects, items that need human judgment, or items that don’t apply. Record the reason for each decision.

After every fix, export the EPUB again and rerun the affected checks. Then retest that file in your target reading systems and with assistive technology.

Keep a review log with tool versions, test settings, findings, unresolved issues, and fixes. Document each step, and limit accessibility claims to what the full evaluation supports.

AI-assisted content review before EPUB testing

Use AI only to clean up draft content before EPUB validation. Interactive eBook Builder, a Notion-based workflow by makaroxitheo, can help with draft-stage content QA through Agent C’s polishing and final content QA check. Notion AI is required and is not included.

This preparation does not replace EPUB validation or human accessibility evaluation. It also does not imply built-in EPUB export.

What does the W3C standard say about evaluation?

EPUB Accessibility 1.1 treats metadata and conformance as two separate things. Every EPUB must carry the accessibility metadata (access modes, features and hazards), but a conformance claim is optional and has extra conditions: the package must carry a dcterms:conformsTo value in the exact pattern the spec defines, such as "EPUB Accessibility 1.1 - WCAG 2.2 Level AA", and it must name the evaluator with a11y:certifiedBy (W3C: EPUB Accessibility 1.1). A tool report cannot supply either condition for you. Someone has to decide that the content meets the stated WCAG level, which is the manual work described above.

In practice, that gives self-publishers two honest paths. If you have not done the full evaluation, publish the descriptive metadata and a plain-language summary of known limits. If you have, record the claim, the evaluator and the date of the review in your log. The metadata fields themselves are explained in our beginner’s guide to EPUB accessibility metadata.

How do you fit this into a release routine?

  1. Export the EPUB from your source project. If the validator reports a structural error such as a spine idref mismatch, fix that first, since structural errors can hide other findings.
  2. Run EPUBCheck, then Ace, and save both reports with the file version.
  3. Do the manual pass on one tablet and one desktop browser.
  4. Write the summary and metadata from what you found.
  5. Repeat after any edit. Keep the log for the next release.

Format and export decisions come before all of this; our comparison of EPUB, MOBI and AZW3 explains why EPUB is the starting point for most workflows.

Final EPUB accessibility checklist

Before release, check these three areas:

After any EPUB change, rerun the relevant automated and manual checks before release.

FAQs

How do I choose an EPUB accessibility conformance target?

Match your content goals to what your publishing tools can support. Check whether your template or workflow allows alt text, captions, high-contrast layouts, and audio controls.

Automated checks alone can’t establish accessibility conformance. Review image descriptions by hand to make sure they convey meaning, and check that the reading order makes sense. Test across devices and browsers to confirm that navigation works as expected and interactive elements stay accessible.

Which reading systems should I test?

Test manually on at least one tablet and one desktop browser. Automated tools alone can’t verify the user experience. Include iPad and Android tablets, plus browsers such as Chrome and Safari, with priority given to the devices your audience uses.

Check audio triggers, hotspot taps, and page navigation under different autoplay rules. Look for tap targets that are hard to hit, low contrast, and layout problems.

How should I prioritize accessibility fixes?

Focus first on barriers that keep readers from moving through and using your content. Add descriptive image alt text, keep the reading order logical across devices, and make interactive elements accessible by keyboard or touch.

Automated checks alone cannot confirm conformance. Manually review image descriptions, reading order, predictable navigation, and caption controls for audio or video. Test early with actual users to find the issues that make reading hardest.

FAQ

How do I choose an EPUB accessibility conformance target?

Match your content goals to what your publishing tools can support. Check whether your template or workflow allows alt text, captions, high-contrast layouts, and audio controls. Automated checks alone can’t establish accessibility conformance. Review image descriptions by hand to make sure they convey meaning, and check that the reading order makes sense. Test across devices and browsers to confirm that navigation works as expected and interactive elements stay accessible.

Which reading systems should I test?

Test manually on at least one tablet and one desktop browser. Automated tools alone can’t verify the user experience. Include [iPad](https://www.apple.com/accessibility/) and [Android](https://www.android.com/accessibility/) tablets, plus browsers such as [Chrome](https://www.google.com/chrome/) and [Safari](https://www.apple.com/safari/), with priority given to the devices your audience uses. Check audio triggers, hotspot taps, and page navigation under different autoplay rules. Look for tap targets that are hard to hit, low contrast, and layout problems.

How should I prioritize accessibility fixes?

Focus first on barriers that keep readers from moving through and using your content. Add descriptive image alt text, keep the reading order logical across devices, and make interactive elements accessible by keyboard or touch. Automated checks alone cannot confirm conformance. Manually review image descriptions, reading order, predictable navigation, and caption controls for audio or video. Test early with actual users to find the issues that make reading hardest.

Ship niche eBooks with agents already built.

Get the Interactive eBook Builder for $59