EPUB Accessibility Checker

This is a first-pass accessibility screen for anyone shipping a reflowable EPUB: a self-publisher, an editor sending files to a library, or a developer checking a conversion. It targets gaps that matter to screen-reader users and can be read from the files: missing alt attributes, skipped heading levels, pages without a language or title, a missing table of contents, and missing accessibility metadata.

Validate the structure first with the EPUB Validator, use this checker for accessibility, then open the EPUB Inspector to find the file a page or image lives in.

Drop a single .epub; it is parsed in your browser and never uploaded. DRM-protected files are refused.

The report groups checks into metadata (book title, language, author, and the schema.org accessibility properties accessMode, accessibilityFeature, accessibilityHazard and accessibilitySummary), navigation (an EPUB 3 navigation document and a table of contents), structure (the mimetype file and the reading order), images, and text. Every page in the spine is opened: images are counted with and without an alt attribute, heading levels are followed to find skips such as an h2 followed by an h4, and each page is checked for a lang or xml:lang declaration and a title. Each line is marked pass, warning or fail, and a text report can be downloaded.

What it does not do: it does not check color contrast (not derivable from EPUB source), does not judge whether existing alt text is meaningful, does not compare the reading order with the visual layout, and does not verify full WCAG 2 or EPUB Accessibility 1.1 conformance.

The biggest trap is reading a clean alt-text result as 'images are described.' The check only confirms an alt attribute exists on each img element in the book's pages; alt="image1" still passes, and alt="" is counted separately because it marks an image as decorative, which is right for an ornament but wrong for a chart. Images drawn with SVG or set as CSS backgrounds are not counted.

Heading checks are per page: a page that starts at h2 is fine, but a jump from h2 to h4 within a page is reported. Pages with no heading are listed as normal for covers and copyright pages; only a book in which no page has a heading is flagged. The accessibility metadata check confirms the four properties are declared, not that their values are true.

A passing report is a floor, not a certificate. The screen-reader experience also depends on a sensible reading order, meaningful link text, table markup, and marked language changes within a page, none of which this checker evaluates.

Fix failures first, such as a missing title or language or images with no alt attribute, then work through the warnings. Write alt text that says what an image shows or why it matters, and use alt="" only for decoration. If the accessibility metadata is missing, add it to the package file; the EPUB Accessibility specification lists the expected values.

Because this is a basic automated screen, re-run it after each fix so you can see counts move, and treat a passing image check as a prompt to review alt-text quality yourself. For anything going to a library, an institutional collection, or a market with legal accessibility requirements, run a full audit with Ace by DAISY and have a person verify navigation and reading order before release.