5 Accessibility Challenges Publishers Keep Running Into (And What Actually Fixes Them)

Walk the accessibility panels at this year's Frankfurter Buchmesse, running October 7 to 11, and you'll hear the same five problems again and again. Not because publishers aren't trying. Most are. The problem is that accessibility work looks simple from a distance and turns messy the moment someone opens the actual files. A scanned backlist title from 2003. A physics textbook with four hundred equations. A picture book where the story lives entirely in the illustration.

The regulatory clock isn't waiting for anyone to figure this out at their own pace. The European Accessibility Act became enforceable in June 2025. In the United States, large public entities already had to meet ADA Title II requirements by April 2026, and smaller public entities face the same deadline in April 2027. Both point back to the same technical target: WCAG 2.1 AA, layered with the EPUB Accessibility 1.1 specification for digital publications.

Knowing the target doesn't make hitting it easy. Here are the five problems that come up on almost every accessibility engagement, and what it actually takes to solve each one.

1. Backlist titles that were never built to be read by a machine

Frontlist content gets built accessible from the start, or close to it. Backlist content is a different story. A scanned PDF from fifteen years ago is often just a stack of page images with no underlying text layer, no tags, no reading order. A screen reader hits it and finds nothing to read.

Fixing this means OCR first, then structural tagging on top: headings, paragraphs, lists, and a reading order that actually matches how a sighted reader would move through the page. That's not a batch job you run overnight. It's page-by-page reconstruction, and doing it at catalogue scale is where most in-house teams stall out. The instinct to remediate everything at once usually backfires. A prioritised roadmap, starting with the highest-demand titles, gets further than trying to fix ten thousand files simultaneously.

2. Math, tables, and anything that isn't a straight paragraph

Plain prose is the easy case. Publishers get into trouble with STEM content, financial tables, and any layout where meaning depends on position rather than sequence. An equation rendered as an image tells a screen reader nothing. A table with merged cells and no header associations reads as a wall of disconnected numbers.

The fix runs through MathML for equations, so the underlying structure survives translation to speech, and through properly coded table markup with row and column headers explicitly tied to their data cells. Neither is optional if the content includes numbers that matter. A finance publisher that skips this step hasn't remediated the file. It's remediated the parts that were easy and left the parts that were actually load bearing.

3. Alt text that has to work at volume, not just in a single sample chapter

Every image needs a text alternative. That rule is simple. What's hard is writing several thousand of them for a single catalogue without either burning weeks of editorial time or producing alt text so generic it's useless. "Image of a chart" tells a reader nothing about what the chart actually shows.

Good alt text is specific to what the image communicates in context. A chart showing rising sales figures needs alt text that conveys the trend, not just the fact that a chart exists. Getting that right at scale takes a workflow that pairs automated first-pass description with human editorial review, because neither one alone gets there. Automation misses nuance. Manual writing alone doesn't scale to a full backlist.

4. Files where the reading order has nothing to do with the visual layout

This one hides well. A page can look perfectly normal to a sighted reader while the underlying code reads captions before body text, sidebars before the main narrative, or footnotes in the middle of a sentence. Nobody catches it by looking at the page. You only catch it by testing with an actual screen reader, or a tool like ACE by DAISY for EPUB files.

Legacy files are the worst offenders here, especially anything converted from print layout software that cared about visual position and nothing about document structure. Getting reading order right means going back to the underlying markup, not just the visual proof, and confirming that assistive technology encounters content in the order intended.

5. Proving it, not just doing it

A publisher can do the remediation work and still fail the compliance conversation if they can't document it. Distributors, procurement teams, and institutional buyers increasingly ask for proof: conformance metadata declaring WCAG level, a certifier's report, an accessibility statement covering what's done and what's still in progress. EPUB Accessibility 1.1 requires exactly this kind of metadata: a conformsTo property naming the WCAG level met, plus information on who certified it. Skipping this step means a publisher can have genuinely accessible files and still lose a deal, because the buyer's compliance team has no way to verify the claim without opening every file individually.

Where this actually gets solved

None of these five problems get fixed by good intentions alone. They get fixed by a workflow built for exactly this kind of work: automated detection that flags issues at scale, paired with human reviewers who catch what automation misses, running against a documented standard from the first pass through final certification.

That's the shape of the remediation pipeline behind S4Carlisle's Benetech GCA certification. The NINJA AI Ecosystem handles first-pass detection and correction across OCR, tagging, alt text drafting, and reading order checks, while trained reviewers verify the output against WCAG 2.1 AA and EPUB Accessibility 1.1 before anything ships. Backlist or frontlist, trade or STM, the pipeline doesn't change shape depending on what kind of file walks in.

S4Carlisle is Benetech GCA certified and runs accessibility remediation at scale through the NINJA AI Ecosystem. Visit us at Frankfurt to talk through where your catalogue stands.