Human Accessibility Testing for EPUB: Why Automated Checks Are Not Enough

An EPUB Can Pass Technical Checks and Still Be Difficult to Use

An EPUB can appear technically correct and still create barriers for a reader using assistive technology.

A heading may exist but sit in the wrong hierarchy. An image may contain alternative text but fail to communicate its meaning. A table may be structurally correct yet still be difficult to understand.

That is why EPUB accessibility testing should not end with an automated check.

The real question is: “Can the reader navigate, understand and use the publication comfortably?”

Technical validation helps identify issues. Human testing helps determine whether the publication actually works in real use.

Why Human Testing Still Matters

Automated accessibility checks are valuable because they can identify repeatable technical problems quickly and consistently. Human accessibility testing looks at areas that require interpretation, context and judgment.

For example, an automated check may confirm that an image has alternative text. A human reviewer can determine whether that description gives the reader the information they actually need.

The stronger approach is:

Technical Validation → Human Review → Correction → Retesting → Final QA

Automation and human testing should support each other, not compete with each other.

What Is EPUB Accessibility Testing?

EPUB accessibility testing evaluates whether a digital publication can be understood, navigated and used by people with different accessibility needs.

A complete review can look at heading hierarchy, reading order, navigation, image descriptions, tables, links, notes, accessibility metadata, complex content and behaviour with assistive technologies.

The purpose is not simply to produce a list of errors.

The purpose is to create an EPUB that readers can actually use.

Automated Testing and Human Testing Play Different Roles

Automation improves speed and consistency.

Human testing adds something different: context, judgment and real-reader insight.

Automation can tell you that something is present. Human testing helps determine whether it actually works.

Where Human Judgment Makes the Biggest Difference

Image Descriptions

Imagine an educational EPUB contains a diagram showing the major parts of the human heart.

The image description says:

“Diagram of a heart.”

Technically, alternative text exists.

But that does not necessarily give a student the information the diagram is meant to teach.

A human reviewer can consider the purpose of the visual, the information the reader needs and whether equivalent information is already available in the surrounding text.

A decorative image may need little or no description. A meaningful diagram, chart or illustration may need much more.

Good alt text communicates purpose and meaning—not simply appearance.

Heading Structure

A publication may contain headings at every important point and still have a confusing hierarchy.

For example, the structure may jump from a chapter directly into a smaller subsection before returning to a main section.

A reader navigating by headings needs to understand how each part of the publication relates to the next.

A clearer structure follows a logical relationship:

Chapter → Section → Subsection

Human QA helps confirm that the technical heading structure matches the actual organisation and meaning of the content.

Tables and Complex Content

Tables can be especially challenging in academic, educational and technical EPUBs.

A table may contain structural markup but still be difficult to understand because its headers are unclear, relationships are complex or meaning depends heavily on visual position.

Human testing asks a practical question:

“Can the reader still understand the information when the visual grid is not the main way of accessing it?”

Accessible structure is important.

Understandable information is the real goal.

Why Assistive-Technology Testing Matters

Assistive technologies can expose an EPUB differently from visual reading.

A publication that looks clear on screen may behave very differently when someone navigates it by headings, links, keyboard commands or a screen reader.

Testing can reveal confusing navigation, unexpected reading sequences, unclear links, weak image descriptions, difficult table relationships and other barriers that may not be obvious during visual review.

For publishers, this means technical conformance and real usability should be treated as connected but separate parts of quality assurance.

A file can satisfy many technical checks and still need usability improvement.

What Do Current EPUB Accessibility Standards Mean for Publishers?

EPUB Accessibility 1.1 is a W3C Recommendation that defines requirements for evaluating EPUB accessibility and making accessibility information discoverable through metadata.

A newer EPUB Accessibility 1.2 specification is currently a Candidate Recommendation Draft dated 18 August 2026.

Because accessibility standards continue to develop, publishers should clearly identify which specification and version their production workflow, customer or contract requires.

The practical lesson is simple:

Standards define the accessibility target. Testing determines whether the publication actually reaches it.

A strong publishing workflow should be able to adapt as standards evolve without rebuilding the entire production process.

Before You Publish: 7 Things Your EPUB QA Should Confirm

Before approving an accessible EPUB, check that:

• Structure is logical — headings, sections, lists and notes follow a clear hierarchy.
• Navigation works — readers can move between chapters, links and notes easily.
• Images communicate meaning — important visuals have useful descriptions.
• Tables remain understandable — relationships are clear without relying only on visual layout.
• Accessibility metadata is complete — available accessibility features are described accurately.
• Human testing is included — the EPUB has been reviewed beyond automated checks.
• Issues are retested — corrections are verified before final approval.

“The goal is not simply to find accessibility issues. It is to make sure the finished ebook works well for real readers.”

A Practical EPUB Accessibility QA Workflow

A strong accessibility workflow does not need to be complicated.

Assess

Understand the publication type, content complexity and accessibility requirements before testing begins.

Validate

Run technical checks to identify structural, metadata and conformance-related issues.

Review

Evaluate headings, navigation, image descriptions, tables, links and other meaningful content.

Test

Check representative reading journeys using assistive technologies and human review.

Correct

Resolve the barriers identified during validation and testing.

Retest

Confirm that corrections solved the original problems without introducing new ones.

Approve

Complete final QA before the EPUB is delivered or published.

The workflow can be summarised as:

Assess → Validate → Review → Test → Correct → Retest → Approve

The most important part is the feedback loop.

Testing without correction is only an audit.

Turn QA Findings Into Better Future EPUBs

Accessibility testing should improve more than one title. Imagine the same navigation issue appears across 40 ebooks.

Correcting those 40 titles solves the immediate problem. But identifying why the issue entered production can prevent the same defect from appearing in the next 400 titles.

That changes the workflow from:

Find → Fix → Forget

to:

Find → Fix → Learn → Prevent

This is where accessibility QA begins improving the wider publishing process. The strongest accessibility programme does not only correct defects.

It reduces the number of defects created in the future.

Human QA Is Especially Important at Scale

When a publisher is working with one EPUB, individual review is relatively straightforward.

When the catalogue contains hundreds or thousands of titles, accessibility QA needs to become a repeatable production process.

Representative sampling, clear defect categories, correction rules and retesting criteria help teams maintain consistency across larger volumes.

The aim should not be to treat every title as a completely separate accessibility project.

Instead, publishers can identify recurring problems, establish consistent remediation rules and feed the findings back into future production.

That creates a more scalable approach to accessible publishing.

How Gentize Innovations Supports Human Accessibility QA

Gentize Innovations supports publishers and digital-content teams with accessible EPUB and EPUB3 production, EPUB remediation, alt-text creation including complex content, screen-reader and keyboard testing, human accessibility QA, accessibility validation and remediation support, structured-content conversion, and publishing production QA.

Gentize has 50+ A11Y-certified testers with visual impairments, helping accessibility evaluation include real assistive-technology experience.

Gentize is also supported by a 200+ member production team, allowing accessibility QA to connect with larger publishing-production programmes.

The focus is not simply on running another technical check.

It is about combining:

Technical Production + Human Judgment + Assistive-Technology Testing + Repeatable QA

For publishers managing larger catalogues, this approach can help accessibility become part of normal production rather than a separate task added at the end.

Final Takeaway

Automated accessibility testing and human accessibility testing answer different questions.

Automated testing asks:

“Can we detect a technical issue?”

Human testing asks:

“Can the reader actually understand and use this?”

The strongest EPUB accessibility testing process combines technical validation, human judgment, assistive-technology testing, correction and retesting.

For one title, that creates a better reading experience.

Across a publishing programme, it creates something even more valuable:

a repeatable accessibility workflow that improves both current titles and future production.

Accessibility should be part of publishing quality—not simply another checkbox before release.