EPUB Accessibility Basics: Metadata, Structure and Checks

An EPUB is a website in a box: a ZIP file of XHTML pages, styles, images and a package document that lists them. Readers who use screen readers, braille displays, large print or text-to-speech depend on how that content is built. This guide explains the main parts of an accessible EPUB and what an automated check can and cannot tell you.

Discovery metadata

EPUB Accessibility 1.1, a W3C Recommendation, asks publishers to describe the accessibility of a book in the package document so that retailers and libraries can show it before someone buys. It uses schema.org properties: schema:accessMode (textual, visual, auditory), schema:accessModeSufficient (the combinations that are enough to read the book, such as "textual"), schema:accessibilityFeature (alternativeText, tableOfContents, structuralNavigation, MathML, pageNavigation, and others, or "none"), schema:accessibilityHazard (none, flashing, motionSimulation, sound, or unknown) and schema:accessibilitySummary, a plain-language paragraph. A conformance statement, dcterms:conformsTo, can claim a level such as "EPUB Accessibility 1.1 - WCAG 2.1 Level AA", optionally with a certifier. The claim is the publisher's, and nothing in the file proves it.

Structure that screen readers use

• Language. dc:language in the package and lang on the content, so the right voice and hyphenation are used.
• Title and identifier. Needed for the book to be listed and found.
• Navigation document. A table of contents with working links is the main way to move through a book; a page list lets readers jump to print page numbers when the book has a print source; landmarks mark parts such as the body and the cover.
• Headings. Real h1 to h6 elements in a sensible order. Bold paragraphs that merely look like headings give no structure.
• Reading order. The spine and the document order are the reading order, so both should follow the logical flow.

Content that needs a text alternative

• Images. Every image needs an alt attribute: text describing it, or an empty alt="" for a purely decorative picture. A missing attribute is an error; an alt text that is a file name is no better. Complex images need longer descriptions.
• Tables. Header cells (th), with scopes in complex tables, and a caption where useful. Layout tables should be avoided or marked as presentation.
• Math. MathML with an alternative text or annotation, not pictures of formulas.
• Audio and video. Captions, subtitles and transcripts.
• Links. Link text that makes sense out of context, not "click here".
• Scripts and forms. If used, they must work with the keyboard and be labelled, and the manifest declares scripted content.

Fixed layout

A fixed-layout EPUB positions content like a printed page, often as images or absolutely placed text. It is harder to make accessible: text can be an image of text, and the reading order can be unclear. Where a fixed layout is needed, provide real text and check the reading order by hand.

What a program can and cannot find

The EPUB Accessibility Checker reads the package, the navigation document and every content document in your browser and reports what a program can detect: missing metadata, language, titles, headings that skip levels, images without alt, tables without header cells, empty and generic links, MathML without alternatives, videos without captions, scripted content and navigation problems, each with the file and the related WCAG success criterion. It also compares the declared features with the content, and it flags a conformance claim that the findings do not support.

It cannot judge reading order, colour contrast, whether alt text is useful, language changes within the text or whether the book makes sense when read aloud. Those need a person with a screen reader and accessibility knowledge. Its findings are technical and separate from legal compliance.

The European Accessibility Act context

The European Accessibility Act (Directive (EU) 2019/882) brought e-books and dedicated reading software into scope from 28 June 2025. How a publisher shows compliance depends on national laws and on harmonised standards, for which EN 301 549 and WCAG are the usual reference. A preflight is a way to find and fix defects early; it is not a conformity assessment. For PDFs, the PDF Accessibility Checker does a similar preflight, and the Screenshot Color Contrast Checker helps with the contrast question the EPUB check cannot answer.