The Definitive Accessibility Guide for PDFs / Office Documents / EPUB: Tagging, Reading Order, Alt Text, and Export

Overview (Key Takeaways First)

• Ensure structure (tagging), reading order, and text alternatives so that PDF, Word, PowerPoint, and EPUB let both “readers” and assistive technologies arrive at the same understanding.
• For images, charts, equations, footnotes, and annotations, it’s crucial to decide what to silence vs. what to describe, based on purpose.
• With proper TOC, headings, and landmarks (EPUB), screen-reader users can jump to targets instantly.
• For PDFs, aim for a tagged PDF; for Word/PowerPoint, use styles and outline; for EPUB, rely on semantic HTML and the nav element.
• Share a publish workflow (manuscript → template → verification → export → re-verification) and checklists across your team to raise reproducibility.

Intended readers (concrete): Corporate comms / HR / admin staff who distribute materials; educators building courseware; research & academic secretariats; municipal/public-sector web managers; publishers / editors / designers; product documentation owners (PMs / technical writers / CS).
Accessibility level: Target WCAG 2.1 AA as the baseline and aim for AAA where feasible (enhanced contrast, improved readability).

1. Why do we need document accessibility? Because focusing on the Web alone still leaves people out.

Even if your web pages are accessible, if your PDFs, Word files, PowerPoints, and EPUBs are hard to read, application forms, lectures, specs, and release notes still become walls that shut someone out.
In public procurement, education, healthcare, research, legal, and finance, accurate comprehension of documents feeds directly into decisions. If assistive technologies (screen readers, braille displays, magnifiers, etc.) can obtain the document structure, readers can reach the “target chapter,” “needed table,” or “relevant note” without getting lost. Without structure, readers face the drudgery of linear reading from the top, widening the information gap.

This guide keeps tool-dependence minimal and explains the principles and workflow that work across tools. You’ll also find templates and checklists you can drop into production tomorrow.

2. First principles: semantics before looks — design with “semantics first”

• Headings as heading elements: Don’t fake it with big bold paragraphs—use heading styles (Word) / heading placeholders (PowerPoint) / <h1>–<h6> (EPUB/HTML).
• Paragraphs as paragraphs; lists as lists: Don’t align with indents alone—use numbered/bulleted lists (Word/PPT list tools; HTML’s <ol> / <ul>).
• Tables as real tables: Don’t lay out with tabs/spaces—use table tools (Word/PPT) or <table> (HTML), and declare header rows/columns.
• Emphasis by meaning: Don’t rely on color/size alone—use bold/emphasis styles. Color should support, not carry, meaning.
• Reading order is logical order: Even with complex layouts/columns, set the reading order (Order pane / outline / DOM order) correctly.

Just following these boosts the information exposed to assistive tech across PDF and EPUB and dramatically reduces misreads.

3. Images, figures, icons, equations: separate what to describe from what to silence

3.1 Designing alternative text (alt)

• High informational value (diagrams, result charts, process flows) → provide a concise summary. Example: “Flowchart showing the five steps of online application: ID check → input → review → payment → completion.”
• Functional icons (buttons) → rely on the button label; keep the image itself silent.
• Decorative images → provide empty alt (artifact in PDF tags; mark as decorative in Word/PPT) to suppress announcement.

3.2 Double-up key points in both figure caption and body

State the figure’s takeaway in the body text and add a caption under the figure. Don’t let critical conclusions live only inside an image.

3.3 Math & chemistry

• EPUB/HTML: Prefer MathML (ideal) or LaTeX rendered with MathJax, keeping a readable structure.
• PDF: If possible, export accessible math with embedded MathML. If not, provide textual expressions or explanatory prose alongside.

4. Tables: relationships are everything — header cells, scope, and summaries

• Declare header rows/columns: Mark the first row as column headers and the first column as row headers.
• Split complex tables: For multi-level/merged headers, split into multiple tables or add explanatory text.
• scope in EPUB/HTML: Use scope="col" for column headers and scope="row" for row headers; for complex tables, use headers + id references.
• Table title/caption: Clearly state purpose; in the body, write key takeaways (max/min, trend).
• Units: Indicate units in cells or header labels (e.g., [cases] [¥] [%]).

5. Headings, TOC, and bookmarks: build the “map” for rapid navigation

• Honor heading levels: One H1 for the document title. Chapters as H2, sections as H3, etc. Change visual size via styles, not levels.
• Auto-generated TOC: Build from heading styles (Word/EPUB). For PDFs, create bookmarks from headings and verify their structure/order.
• TOC with page numbers alone is unfriendly: Screen-reader users don’t navigate by page numbers. Make TOC entries clickable links to jump.

6. Reading order & tab order: align visual layout with the “audio order”

• PowerPoint: On each slide, verify the reading order (Selection pane) and reorder to Title → body → figures → notes.
• PDF: Align the tag tree and reading order panels; keep logical order even with irregular frames/columns.
• EPUB/HTML: DOM order equals reading order. Use CSS reordering (order, position) sparingly.

What feels natural visually can sound scrambled—e.g., “title, then figure note first, body later.” Order checks make or break final quality.

7. Contrast, fonts, and spacing: the basics of readability

• Text contrast: ≥ 4.5:1 for normal text; ≥ 3:1 for large text. If text overlays photos, add solid underlays or shadows.
• Fonts: Choose high-legibility faces for body text. If headings are decorative, ensure adequate weight/size/leading.
• Line length & spacing: Aim for 40–80 characters per line; ~1.5 line spacing. Use generous paragraph spacing to separate blocks.
• Don’t rely on color alone: Reinforce instructions/warnings with text/shape/symbols, not “red text = important.”

8. Notes, footnotes, references: provide destinations and return paths

• Bidirectional links: Link body → note and note → body (PDF bookmarks/links; EPUB anchors).
• Brief synopsis first: When a footnote is long, put a short summary up front.
• References: Link citations to in-text locations; include a bibliography. Keep sources as text, not images.

9. PDFs with forms (applications / fillable sheets): input assistance and error communication

• Labels for fields: Keep visual labels and programmatic names consistent.
• Tab order: Follow a logical input sequence (Name → Contact → Address → Review → Submit), not just top-left to bottom-right.
• Required vs optional: Indicate in text (e.g., “*Required”). Don’t rely on color; place notes close to fields.
• Errors: After validation, show a summary at the top plus field-level messages.
• Signature fields: Add instructional text. If print-and-sign is an option, state signature choices clearly.

10. EPUB specifics: semantic HTML and navigation

• Document skeleton: Use <main>, <section>, <article>, <nav> appropriately.
• Heading continuity: Start chapters with <h1> or <h2> consistently; keep internal levels coherent.
• TOC (nav): Provide a structured TOC via <nav epub:type="toc">.
• Image alternatives: Provide alt and, when needed, explain in body text to support comprehension.
• Metadata: Accurately set title, author, language, and identifiers; don’t forget <html lang="ja"> for Japanese.

11. Workflow: Manuscript → Template → Structuring → Verification → Export → Re-verification

1. Manuscript
   • Sketch chaptering, heading levels, figure/table placement, and footnotes on paper—as a storyboard.
2. Template
   • Word: Style guide for Heading 1–3, Body, Quote, Figure/Table captions.
   • PowerPoint: Accessible master layouts (Title+Body; Title+Two Columns; Figure+Caption).
   • EPUB/HTML: Semantic boilerplate and common CSS (contrast, spacing, font sizes).
3. Structuring (authoring)
   • Apply styles for meaning, add alt text, declare table headers, clarify link text.
4. First pass checks
   • Automated checks (missing styles, heading skips, absent alt text) plus manual reading-order checks.
5. Export
   • Export PDFs as tagged PDFs. Validate EPUBs with a validator.
6. Re-verification (accessibility)
   • With NVDA/VoiceOver, test navigation TOC → chapter → table → footnote.
   • At 150–200% zoom, confirm reflow and readability.
7. Release & replacement operations
   • Put version and update date in filename and document. When replacing, state where the old version is archived.

12. Case study: Turning a municipal “guideline PDF” into a readable document

Before

• Nicely typeset PDF, but text is images, not searchable.
• Headings are big bold paragraphs; no bookmarks.
• Tables aligned with tabs; reading order is chaotic.

After

• Re-authored in Word with heading styles; auto-generated TOC.
• Tables rebuilt with table tools; column headers set to “Category / Amount / Conditions.”
• Notes summarized in body text.
• Exported as a tagged PDF; bookmarks generated from headings.
• Verified in NVDA: heading jump, header announcement in tables, footnote jump/return.
• Result: Fewer misunderstanding inquiries; time-to-targeted section halved from search traffic.

13. Common pitfalls and how to avoid them

14. Sample: The fastest Word → Tagged PDF template

Word styles (example)

• Heading 1: Document title (first page only)
• Heading 2: Chapter titles (numbered)
• Heading 3: Section titles
• Body: legible Japanese fonts (e.g., Yu Gothic/Mincho), 12 pt, 1.5 line spacing
• Caption: “Figure n / Table n” (auto numbering)
• Bulleted/numbered lists: defaults (tighten indents slightly)

Creation steps (highlights)

1. Enter Title → Heading 2 → Body in order.
2. Add figure captions and table titles before referencing them in text.
3. When inserting images, add alt text (mark decorative images as decorative).
4. Insert auto TOC.
5. Run Accessibility Checker (if available) and resolve warnings.
6. Save → Export as PDF (check Tagged PDF).
7. In the PDF, verify bookmarks & reading order (tweak if needed).

15. Sample: Cleaning up PowerPoint “reading order”

• Define placeholders and order for Title, Body, Figures, Captions in Slide Master.
• On each slide, open the Selection pane and reorder so top to bottom = reading order.
• Add alt text to figures; mark decorative items hidden from reading.
• Avoid scattered text boxes; express logic with lists and tables.
• Create slide-internal links (TOC slide → chapter slides) for rapid navigation.

16. Sample: EPUB/HTML skeleton (minimal semantics)

<!doctype html>
<html lang="ja">
<head>
  <meta charset="utf-8">
  <title>アクセシブルEPUBの基本</title>
  <meta name="viewport" content="width=device-width,initial-scale=1">
</head>
<body>
  <header>
    <h1>アクセシブルEPUBの基本</h1>
    <nav aria-label="目次">
      <ol>
        <li><a href="#ch1">第1章 はじめに</a></li>
        <li><a href="#ch2">第2章 画像と代替</a></li>
        <li><a href="#ch3">第3章 表と見出し</a></li>
      </ol>
    </nav>
  </header>
  <main id="main">
    <article id="ch1">
      <h2>第1章 はじめに</h2>
      <p>この章では…</p>
    </article>
    <article id="ch2">
      <h2>第2章 画像と代替</h2>
      <figure>
        <img src="flow.png" alt="申請の流れ：登録→入力→確認→送信→完了">
        <figcaption>図1 申請の流れ</figcaption>
      </figure>
    </article>
    <article id="ch3">
      <h2>第3章 表と見出し</h2>
      <table>
        <caption>手当の一覧</caption>
        <thead><tr><th scope="col">区分</th><th scope="col">金額</th></tr></thead>
        <tbody><tr><th scope="row">A</th><td>10,000円</td></tr></tbody>
      </table>
    </article>
  </main>
  <footer>© 2025 Example</footer>
</body>
</html>

17. Verification: combine automated checks with a short manual pass to guarantee “actually readable”

Automated checks (common criteria)

• Presence of alt text (including silencing decorative items)
• Heading level skips (e.g., H1 → H3)
• Clarity of link text (avoid “click here”)
• Color-only reliance (is meaning conveyed only by color?)
• Existence of PDF tags / anomalous reading-order patterns

Manual 5-minute routine

1. Can you jump from TOC → chapter instantly?
2. In a screen reader, do the headings list and links list make sense?
3. In tables, are headers announced during navigation?
4. Do figures stay silent or summarize appropriately?
5. At 150% zoom, do lines avoid overlap/cutoff?
6. In PDFs, do bookmarks & reading order reflect logical order?

18. Rolling out in organizations: templates, training, and audits make it sustainable

• Template distribution: Place Word templates (with styles), accessible PPT layouts, and EPUB skeletons on a shared drive.
• Training: Run recurring 15-minute micro-sessions (alt-text triage / heading levels / reading order).
• Audits: Quarterly, sample 5 docs at random and review; fold improvements into knowledge base and templates.
• Definition of Done (pre-external release):
  • [ ] Heading levels and TOC are correct
  • [ ] Proper alt text for images (decoratives silenced)
  • [ ] Tables have header cells and captions
  • [ ] Reading order is logical
  • [ ] Contrast meets standards
  • [ ] PDF is tagged; EPUB is validated
  • [ ] Version and update date are stated

19. Who benefits, and how? (concrete impact)

• Assistive-tech users: Heading jumps/TOC/bookmarks provide direct access to targets; alt text conveys figure takeaways.
• Older adults / low vision: Reflow-safe layouts and solid contrast reduce fatigue.
• People with diverse cognitive/learning profiles: Headings/lists/summaries lower comprehension and memory load.
• Localization & reuse teams: Semantics strengthen MT, re-layout, and search, boosting in-house efficiency.
• Legal & compliance: Meeting requirements like reasonable accommodation and not conveying meaning by color alone cuts risk.

20. Mini Q&A

Q1. Our PR magazine is image-heavy—what should we do?
A. Provide a text summary page first, giving key facts, dates, contacts per article. Add informational supplements in captions and keep decoration silent.

Q2. We only have scanned PDFs of older docs.
A. OCR them, and at least rebuild the key pages. If that’s not immediate, add contact info and explain how to request alternate formats.

Q3. Our paper is full of equations.
A. If possible, consider embedding MathML. If not, bolster textual definitions/descriptions and end-of-chapter summaries.

21. Summary: Turn documents from “images” back into “words”

1. Semantics first: Express headings, paragraphs, lists, and tables with meaningful tools.
2. Alt-text triage: Decide what to describe and what to silence; keep it focused yet sufficient.
3. Reading order = logical order: Don’t be fooled by visual layout; ensure it sounds natural.
4. TOC, bookmarks, and nav: Enable one-hop travel—the longer the doc, the more crucial the “map.”
5. Workflow & templates: Make it a system, not a personal craft.
6. Habitual quick checks: Pair automation with a 5-minute manual pass for steady quality on every release.

Documents are vessels for words that adapt to readers’ time, energy, and context. With small tweaks, you can make them kinder to many. May your next document let everyone find, read, and understand—without getting lost or exhausted.

22. Accessibility level: evaluation and target

• This guide’s target: WCAG 2.1 AA compliance
  • 1.1.1 Non-text Content (image alternatives)
  • 1.3.1 Info & Relationships (structure of headings/tables/lists)
  • 1.3.2 Meaningful Sequence (reading order)
  • 1.4.1 Use of Color (avoid color-only meaning)
  • 1.4.3 / 1.4.11 Contrast (text / non-text)
  • 2.1.1 Keyboard (PDF forms)
  • 2.4.1 Bypass Blocks (TOC/bookmarks)
  • 2.4.6 Headings & Labels (clear naming)
  • 3.3.x Input Assistance (required indicators; identifying and suggesting fixes for errors)
  • 4.1.2 Name, Role, Value (semantics in EPUB/HTML)

Deeper impact by audience

• Municipal/public sector: Notices/calls/applications reach the same understanding, reducing inquiries and counter load.
• Education & research: Courseware and papers become assets that withstand search, read-aloud, and re-editing, expanding learning access.
• Corporate comms/admin: Distributed materials improve continuously, lifting brand trust.
• Publishers/editors: Multi-output (EPUB/print/Web) gets easier, cutting production costs.

23. Appendix: “Quick Checklist” to tape on the wall before you start

• [ ] Always use styles (Headings, Body, Lists, Captions)
• [ ] Tables built with table tools; header cells + caption present
• [ ] Image alts: informational → summary / decorative → silence
• [ ] TOC/bookmarks/link text use purposeful labels
• [ ] Reading order is Title → Body → Figures → Notes (logical)
• [ ] Contrast, fonts, and spacing meet readability standards
• [ ] PDF exported as tagged; EPUB validated
• [ ] Run the 5-minute manual check (SR headings list, table reading, zoom test)
• [ ] Version, update date, and contact are clearly stated

Even just these will transform the reading experience. Start small—and keep going steadily.