7 Formatting Mistakes Word and Scrivener Authors Make Before Converting to EPUB

In Word and Scrivener, pressing Enter twice to add a visual gap between sections produces blank paragraphs in the document. When exported to EPUB, those blank lines become empty <p></p> tags that render inconsistently across e-reader devices — some show no gap, others show a large blank space that wasn't in the original manuscript. Paragraph and section spacing belongs in CSS margin-bottom on a defined paragraph style, applied at export time, not in blank lines inside the manuscript text.

Scrivener's compile system uses a structural template for each chapter, then copies that template for every section. If the template contains id attributes on heading or structural elements, Scrivener copies those IDs into every chapter — producing duplicate IDs across the whole EPUB. The compile preview looks correct, Scrivener's own export validation passes, and EPUBCheck catches it as OPF-030 on the first platform submission.

When you paste a chart from Excel or a diagram from PowerPoint into a Word document, it embeds as an EMF (Enhanced Metafile) vector image internally — a Windows-only format that web standards and EPUB don't recognize. EPUB converters either skip EMF images silently or flag a validation error requiring a fallback. Re-saving the chart as a PNG before pasting into Word prevents this from reaching the EPUB at all.

Adding a custom font in Word or Scrivener and exporting to EPUB embeds the font file in the EPUB package. Most commercial fonts — including many default Windows fonts like Cambria and Calibri — prohibit redistribution in embedded files. The converter includes the font because it's in the document; EPUBCheck flags it because the font's license doesn't allow embedding in a distributed file. Switching to a free, embedding-permitted font (any Google Font) before export prevents this class of error entirely.

Scrivener's compile settings include an EPUB version toggle that defaults to EPUB 2 in older project templates. Authors who leave it on EPUB 2 and submit to platforms running EPUB 3 validation see opf:role attribute errors — an EPUB 2 metadata format that EPUBCheck 5.x flags as invalid in files declaring EPUB 3. Switching the Scrivener compile target to EPUB 3 removes this entire class of error, and Scrivener's EPUB 3 output is generally cleaner for current platform requirements.

Authors who rename their exported EPUB's chapter files for organization — "chapter-01.xhtml" to "part-1-opening.xhtml" — without updating the OPF manifest leave dangling references. The manifest declares the original filename; the file now has a different name; EPUBCheck reports RSC-007. If you need to rename files inside your EPUB package, update the OPF manifest item's href to match the new filename before repackaging.

Word and Scrivener both produce EPUB output that looks correct in their own preview panes. Neither tool runs EPUBCheck before export, and neither one tells you whether the resulting file will pass Apple Books Connect's ingestion, Kobo Writing Life's KEPUB conversion, or KDP's manifest checks. The gap between "looks right in Scrivener's compile preview" and "passes platform validation" is where most export-workflow rejections originate. Running the exported EPUB through EPUBCheck before submitting to any platform catches the structural issues that source-tool previews don't surface.