Manuscript to EPUB: What Actually Breaks in Ebook Conversion

Most word processors have an "export to EPUB" button, and for a simple text-only book it sometimes works. For everything else, the gap between a finished manuscript and a file a store will accept and a reader will enjoy is wider than it looks. Here is where conversions actually go wrong, and what a clean one has to get right.

Direct formatting vs. real styles

The single biggest source of messy ebooks is a manuscript that looks formatted but isn't structured. If chapter titles are just "big bold text" a writer set by hand, a converter has no reliable way to know they are headings — so the table of contents comes out wrong, navigation breaks, and reflow behaves unpredictably. A manuscript that uses named paragraph styles ("Heading 1", "Chapter Title", "Block Quote") converts cleanly because the structure is explicit. Fixing this after the fact means re-tagging the whole book.

The things that quietly corrupt a file

• Smart quotes, dashes and ellipses — inconsistent typographic characters, or straight quotes mixed with curly ones, are obvious on an e-ink screen even when they were invisible in the draft.
• Images — screenshots dropped in at 96 DPI look fine on a laptop and blurry on a tablet; images wider than the content area get clipped; missing alt text fails accessibility checks some stores now enforce.
• Notes — footnotes and endnotes need to become linked, round-trippable references, not raw superscript numbers that go nowhere.
• Front and back matter — title page, copyright page, dedication and "also by this author" pages each need to be their own navigable section, not run together.
• Embedded fonts — a decorative font that isn't licensed for embedding, or isn't embedded at all, silently falls back to something else on the reader's device.

What a clean EPUB actually contains

A valid EPUB is a small website in a box: XHTML content files, a navigation document that drives the reader's table of contents, a package file that lists everything in reading order, and metadata (title, author, language, identifier). "Good" means it also has semantic headings in the right order, alt text on images, a logical section break per chapter, and no validation errors — because a file that fails EPUBCheck can be rejected at upload or render inconsistently across apps.

What conversion can't do for you

Automation turns a well-structured manuscript into a clean, valid ebook quickly and consistently. It can't make structural decisions you haven't made — whether something is a sidebar or a block quote, where a section should break, what belongs in front matter. And it isn't an editor: it won't catch an inconsistent character name or a chapter that runs long. The teams that get a clean ebook on the first pass settle the structure in the manuscript, then let the conversion be mechanical.