A reliable DOCX to EPUB workflow has five stages: clean the Word file so it carries structure, convert it, validate the EPUB, test accessibility and reading order, then check it in the stores' own previewers before upload. Skipping the middle stages is why so many authors get a file that opens but looks wrong, or one a store rejects.
This guide walks through each stage in order, with the checks to run and the mistakes that cause most failures.
Key takeaways
- Reflowable EPUB expresses structure (headings, styles, lists, links), not page appearance. Fix the structure in Word before you convert.
- Treat conversion as a loop: convert, run EPUBCheck, fix, convert again. A one-click export is a first draft.
- Validators catch structural defects. A human still has to judge reading order, image descriptions and how the book looks on a device.
- Accessibility metadata is required in the package document of every EPUB, not an optional extra.
- Never unzip and re-zip an EPUB with an ordinary ZIP tool. Fix the source and re-export instead.
Table of Contents
- Why Word files break when they become EPUBs
- Stage 1: Prepare the DOCX
- Stage 2: Convert
- Stage 3: Validate with EPUBCheck
- Stage 4: Test accessibility and reading order
- Stage 5: Preview in the stores' own tools
- A repeatable checklist
- Our perspective on this workflow
- Where Alhora fits
- FAQ
- Sources
Why Word files break when they become EPUBs
A Word document is built around a page. You see a fixed trim size, manual page breaks, tabs for indents, and spacing made with the Enter key. An EPUB has no fixed page. According to the W3C EPUB 3 overview, reading systems re-render reflowable content dynamically, and readers can change typography. Whatever you positioned by eye gets reflowed, or lost.
The EPUB 3.3 specification makes the same point from the other side. Reflowable layout is the default, and fixed layout is for content that genuinely needs pixel-precise positioning, such as image-led or highly designed books. For a standard novel or nonfiction title, use reflowable. So a good DOCX for conversion is one that says what each paragraph is, not how it looks.
In practice that means:
- Headings, paragraph styles, lists and links matter more than manual tabs, repeated spaces or visual positioning.
- Direct formatting (selecting text and changing the font by hand) is noise. Styles are signal.
- Anything that depends on page numbers, such as "see page 112", has to become a link or be rewritten.
Stage 1: Prepare the DOCX
Spend most of your time here. Every hour in Word saves several in debugging.
Apply real styles
- Use Heading 1 for chapter titles and Heading 2 for scene or section titles. These become the navigation entries readers use to jump around.
- Use one Normal or Body Text style for all body paragraphs. Set the first-line indent in the style, not with tabs.
- Create named styles for block quotes, epigraphs, letters and similar elements, rather than formatting them by hand.
- Use Word's list tools for bulleted and numbered lists, not typed hyphens or numbers.
Remove the layout hacks
- Search for double spaces and manual tabs, and delete them.
- Replace empty paragraphs used as spacing with paragraph spacing in the style.
- Replace manual page breaks with "Page break before" on the chapter heading style.
- Turn on Show Formatting Marks (the pilcrow button) and scan the whole file once. Stray marks show up immediately.
Handle front and back matter on purpose
Title page, copyright page, dedication, acknowledgments and "also by" pages each need to be their own clearly marked section. Decide now which ones belong in the table of contents and which belong in neither the contents nor the reading flow's main path. The landmarks guide explains how EPUB navigation semantics mark the start of the text, the cover and the table of contents so reading systems can find them.
Deal with images, footnotes and special content
- Insert images inline, not as floating objects, and write alt text for every informative image. The EPUB Accessibility specification asks for meaningful alternative text on informative images.
- Keep footnotes and endnotes as Word's native notes. Current converters generally handle those better than hand-typed superscripts.
- If your book has equations, plan for them early. See our guide to MathML in EPUB for standards and fallbacks.
Set metadata before you export
Fill in the title, author name, language and description in the document properties. You will also want an ISBN decision (your own or the store's free identifier) settled before the final export, because the identifier lives in the EPUB's package metadata.
Stage 2: Convert
You have three broad routes, and each trades control for convenience.
| Route | Best for | Main risk |
|---|---|---|
| Free converters (online or Calibre, Pandoc) | Quick drafts, reading copies, technical users | Messy CSS, weak typography, limited control over front matter and navigation |
| Store upload of the DOCX | Simple text-only books | You do not control the output and previews can differ from the final |
| Dedicated book-formatting software | Books you will sell and revise | Needs a license, and a small learning curve |
People in author communities often recommend Calibre for a quick "drag, drop and convert" pass. It works, and it is fine for a proof copy. The usual complaint is that the output is only mostly clean, so you end up editing the result by hand. That is exactly the step that causes trouble later.
What a conversion must produce
Whatever tool you use, the result must be a valid EPUB. According to the EPUB 3.3 specification, that means a package document with metadata, a manifest of resources, a spine that defines reading order, a navigation document, and a correctly built container. If your tool does not give you all of these, you have not finished converting.
Do not hand-edit the ZIP
It is tempting to rename the .epub to .zip, fix a line of XHTML and zip it back up. Don't. The Ace SMART user guide warns that ordinary ZIP programs can produce an invalid EPUB, because the mimetype file has to be packaged in a specific way and order. Fix the source document or the tool settings, then export again.
Stage 3: Validate with EPUBCheck
Run EPUBCheck on every export. It is an open-source W3C project, available as a command-line tool or Java library, and it checks EPUB 2 and EPUB 3 conformance, including problems that can stop a file from opening or rendering properly. The DAISY knowledge base describes it as a free command-line validator that reports anything violating the standard or likely to cause trouble.
The command is short:
epubcheck my-book.epub
Read the output top to bottom. Errors must be fixed. Warnings deserve a decision, not a shrug. Common findings from Word conversions include:
- Missing or malformed navigation document.
- Images referenced in the manifest but absent from the package, or the reverse.
- Invalid or empty heading elements created by blank styled paragraphs.
- Duplicate IDs from repeated bookmarks or anchors.
- Missing language or identifier metadata.
For a longer walkthrough of typical messages and what causes them, read how to fix EPUBCheck errors and pass store validation. The fix is nearly always in the Word source: clean the style, delete the empty paragraph, re-export.
Stage 4: Test accessibility and reading order
A clean EPUBCheck report is necessary, not sufficient. The W3C notes that a clean result is a good starting point for accessibility testing, but it does not replace testing or human review.
The DAISY validation process lays out a sensible sequence: EPUBCheck first, then automated accessibility inspection with Ace, then manual inspection with Ace SMART. Automated tools cannot judge whether an image description is meaningful or whether the reading experience is usable.
The minimum accessibility pass
- Metadata. All EPUB publications must include schema.org accessibility metadata in the package document, such as access modes and an accessibility summary, per the EPUB Accessibility 1.1 specification. A newer 1.2 draft repeats that requirement.
- Alt text. Every informative image gets a real description. Decorative images are marked as such.
- Navigation. Check that the table of contents matches your heading structure, and that every entry goes where it says.
- Reading order. Read the spine order against the order you intend, especially with sidebars, epigraphs and notes.
- Manual check. Open the file with a screen reader or at least with large type and a different font, and see what happens.
Page navigation helps in reflowable books, but pagination changes with viewport size and user font settings, so treat page-list markers as a convenience rather than a guarantee. For a fuller checklist, see our EPUB accessibility checklist built on WCAG and the guide to WCAG EPUB compliance.
Stage 5: Preview in the stores' own tools
Valid does not mean pretty everywhere. Each store renders EPUBs with its own engine and applies its own checks.
- Kobo: test layout and accessibility before upload. Our Kobo EPUB formatting guide covers what to look at.
- Google Play Books: preflight against its requirements and pass EPUBCheck first. See Google Play Books EPUB requirements.
- Amazon KDP: use the previewer Amazon provides, and check that your cover and metadata are consistent with the file.
Look at the book on a phone-sized screen, a tablet-sized screen and a dark-mode setting. Check chapter openings, drop caps, scene breaks, block quotes and the first and last pages. Then change the font and size and check again.
A repeatable checklist
Print this and run it for every release, including small corrections.
- Styles applied, no manual tabs or empty paragraphs.
- Front and back matter in separate, labeled sections.
- Alt text on every informative image.
- Metadata complete: title, author, language, description, identifier.
- Convert.
- EPUBCheck: zero errors, warnings reviewed.
- Accessibility metadata present, navigation and reading order checked by a person.
- Store previews on at least two device sizes.
- Keep the DOCX and the final EPUB together, labeled with a version number.
Our perspective on this workflow
We build book-formatting software, so we see where the DOCX route hurts. The problem is rarely the conversion itself. It is that the feedback arrives late: you export, upload, wait, and a store tells you something is wrong with a file you stopped looking at days ago.
That is why we designed Alhora around real-time validation against publishing standards. An error you see while you are still writing a chapter takes seconds to fix. The same error found after upload takes an afternoon. Accessible EPUB export in Alhora is audit-gated, so the file does not go out until it passes the checks, which keeps the validation loop above from being optional.
We also hold a firm line on AI. The AI assistant in Alhora, Inkia, flags issues and helps with proofreading, but it never writes or alters your creative content. Your words stay yours, and formatting should never be a place where that is in question.
If you format one book once, a free converter plus EPUBCheck can be enough. If you publish series, revise often or format for clients, the loop becomes the job, and tooling that does the loop for you pays back quickly.
— James
Where Alhora fits
Alhora is a book formatting studio for Mac, Windows, Linux and ChromeOS. It turns manuscripts into typeset ebooks and print-ready PDFs, checks them against publishing standards as you work, and exports for KDP, IngramSpark, Draft2Digital, Apple Books, Kobo and Google Play. It also handles batch series formatting, metadata and ISBN management, and version control.
The license is £99, one-off. See the ebook formatting software page for how EPUB export works across stores, or the features overview for typesetting and print. If you are still comparing options, how to choose book formatting software lays out the criteria.
FAQ
Can I just upload my DOCX to a store instead of making an EPUB?
Some stores accept DOCX and convert it for you. The catch is that you do not control the conversion, and what you see in the preview can differ from your intent. An EPUB you built and validated yourself gives you a file you can check, keep and reuse on every store.
Is Calibre good enough for DOCX to EPUB?
For a reading copy or a simple text-only book, often yes. It is free and widely used. For a book you will sell, you should still run EPUBCheck on the result and expect to clean up styles, front matter and navigation yourself.
Why does my EPUB pass EPUBCheck but look wrong?
EPUBCheck tests conformance to the standard, not design quality. A file can be valid and still have odd spacing, missing drop caps or a confusing contents list. That is why the DAISY guidance pairs automated tools with human review. Check the book on real devices.
Do I need accessibility metadata if I am a self-publisher?
Yes, as a matter of the standard. The EPUB Accessibility specification says all EPUB publications must include accessibility metadata in the package document. It tells stores and readers what to expect from the book. Claiming full conformance is a separate step that requires testing.
Should I choose fixed layout to keep my Word design?
For most novels and nonfiction, no. Reflowable is the default and lets readers adjust type to their needs. Fixed layout makes sense for image-led or highly designed books where exact positioning is part of the content.
Sources
- EPUB 3 Overview (W3C)
- EPUB 3.3 (W3C)
- EPUBCheck (W3C)
- EPUBCheck, Accessible Publishing Knowledge Base (DAISY)
- Validation Process, Accessible Publishing Knowledge Base (DAISY)
- Overview, Ace SMART (DAISY)
- EPUB Accessibility 1.1 (W3C)
- EPUB Accessibility 1.2 (W3C)