The most damaging eBook conversion mistakes are rarely cosmetic. They begin with an unreliable source file, weak document structure, the wrong output format, inaccessible content, mismatched metadata, or testing that stops as soon as the book opens.
A publish-ready eBook should reflow cleanly, preserve the intended reading order, provide working navigation, remain understandable with assistive technology, pass current validation, and survive retailer processing. The 12 mistakes below explain what fails, why it happens, and how to fix it without turning conversion into an unnecessarily long technical project.
Before diagnosing a file, use the right baseline:
These dates are useful production facts, not ranking decoration. Recheck retailer and validator requirements immediately before publication because platform rules and software versions change.
The conversion can only be as reliable as its source. A clean DOCX or structured authoring file usually preserves paragraphs, headings, lists, links, and notes better than a print PDF. A PDF describes where content sits on a page; it does not necessarily preserve the logical structure an eBook needs.
Scanned PDFs create another problem. Even when the text is selectable, the OCR layer may contain missing words, wrong characters, broken paragraphs, or an incorrect reading order.
If the manuscript still needs preparation, follow a consistent eBook manuscript preparation process before generating the EPUB.
A reflowable eBook does not have stable page dimensions. Readers can change font size, margins, orientation, and device. Forced page breaks, running headers, page numbers, multi-column layouts, and text positioned with spaces may look acceptable in one preview but collapse on another screen.
Remove print-only elements and encode meaning rather than page geometry. Use real headings, paragraphs, lists, quotations, image captions, and section breaks. Keep intentional page references only when the publication has a reliable page-list strategy or must align with a print edition.
Test at a narrow phone width and at enlarged text. If content overlaps, clips, disappears, or requires horizontal scrolling without a strong reason, the reflow is not finished.
Tabs, repeated spaces, empty paragraphs, manual numbering, and direct font changes create fragile eBooks. They may also prevent assistive technology from recognizing headings, lists, quotations, and navigation.
Apply document styles before conversion and use semantic HTML in the EPUB:
Keep the CSS small and predictable. A clean style system is easier to test across platforms than hundreds of one-off rules.
MOBI is no longer a sensible publishing master, and PDF is not a universal eBook format. For most text-led books, a validated reflowable EPUB 3 is the practical distribution master. Fixed layout is appropriate only when the position of text and images is essential, such as some children’s books, comics, or highly designed educational titles.
Choose the format from the content and distribution plan:
For channel-specific decisions, see how to format an eBook for major platforms.
A visual table of contents is not enough. Readers need functional navigation, and assistive technology needs a logical reading sequence. Common defects include missing chapters, duplicate links, headings skipped in the hierarchy, inaccessible sidebars, and footnotes that do not return to the text.
Open the EPUB with a keyboard and screen reader as well as a mouse or touchscreen. The order should remain understandable without relying on visual position.
Images can appear blurry, oversized, stretched, incorrectly cropped, or detached from their captions. Decorative images may be announced unnecessarily, while meaningful figures may have no alternative text at all.
Use the best licensed source image, preserve its aspect ratio, and size it for the intended display rather than enlarging a small file. Add concise alternative text that communicates the image’s purpose in context. Mark decorative images appropriately so they do not create noise for screen-reader users.
Complex charts may require a nearby explanation or longer description. Do not force all visible labels into one unhelpful alt-text sentence. Check sharpness, caption association, zoom behavior, dark mode, and retailer recompression.
Flattening complex content into an image may preserve appearance but removes semantics and often becomes unreadable on a phone. It also prevents text resizing, search, selection, and effective assistive-technology access.
Simplify or split wide tables, preserve true row and column headers, and test at narrow widths. Encode math accessibly when the target reading systems support it, and provide an equivalent explanation when they do not. Place sidebars in a logical reading position instead of relying on floating page geometry.
If an image is unavoidable, include a structured text alternative and test whether a reader can obtain the same information without seeing it.
Font problems cause missing characters, empty boxes, unexpectedly large files, and inconsistent rendering. A font may also prohibit eBook embedding even when it was licensed for print.
Never convert ordinary body text to images simply to preserve a font choice.
Automatic conversion can strip hyperlinks, send readers to the wrong chapter, duplicate anchors, or break note return paths. External links can also become outdated or unsafe between manuscript delivery and publication.
Test every internal link and a representative set of external links. Make link text descriptive enough to make sense out of context. Preserve note relationships and verify that readers can return to the original sentence. Remove tracking parameters or accidental editing URLs that should not appear in the published book.
Do not make essential understanding depend on a website remaining available. The eBook should still make sense if an external resource later disappears.
Accessibility cannot be repaired fully by adding alt text just before upload. Heading order, reading sequence, language metadata, link purpose, color contrast, tables, notes, and navigation all depend on earlier editorial and structural decisions.
Build accessibility into the source and conversion workflow:
Run an automated checker such as Ace by DAISY, then complete manual checks. Automated tools can identify many defects, but they cannot decide whether reading order or alternative text communicates the intended meaning.
The EPUB package, cover, copyright page, distributor feed, and retailer listing should agree. Conflicts in title, author name, language, publication date, edition, or ISBN can create duplicate listings and attach the wrong file to a product record.
Create one approved metadata sheet and use it across every channel. Confirm:
Do not reuse a print ISBN for an eBook. The need for an eBook ISBN depends on the product and distribution plan; the guide to getting an ISBN for a book explains the format and edition rules.
An EPUB that opens in one app is not necessarily publish-ready. EPUBCheck tests specification conformance, not editorial completeness, visual quality, meaningful alternative text, or retailer processing.
Use three layers of QA:
Retest navigation, images, notes, metadata, and typography after every material repair. Archive the final source, EPUB, validation reports, metadata sheet, and release date so the published version can be traced.
Before release, confirm that:
For books with damaged PDFs, complex layouts, tables, math, multiple languages, or strict accessibility requirements, professional eBook conversion services can reduce rework. The provider should explain its output specification, validation process, accessibility scope, testing environments, correction policy, and exact deliverables.
Starting with an unreliable source without assessing it. If text, reading order, images, notes, or tables are already damaged, automatic conversion usually carries those defects into the EPUB.
No. EPUBCheck detects conformance problems. It does not prove that the book is complete, readable, accessible, visually correct, or compatible with every target store. Manual and retailer testing are still required.
Yes, but one-click output is rarely publish-ready when the PDF contains columns, scans, running headers, footnotes, tables, or complex design. Reflowable EPUB often requires text cleanup and structural reconstruction.
No. KDP ended MOBI upload support for reflowable books in 2021 and fixed-layout books in March 2025. Use one of KDP’s currently supported formats, such as EPUB, DOCX, or KPF, as appropriate.
No. Reflowable EPUB is the better choice for most text-led books because readers can resize text and adapt it to their screens. Fixed layout is for content whose meaning depends on precise placement.
A separately available digital publication may need its own ISBN under the chosen distribution plan. Do not reuse the print ISBN. Some proprietary platforms use their own identifier and do not require an ISBN for the eBook.
There is no useful universal number. Test the target ecosystems and high-risk conditions: narrow screen, enlarged text, relevant retailer preview, standards-oriented reader, and assistive technology. Add fixed-layout, math, media, or right-to-left tests when the book uses those features.
Reliable eBook conversion is a controlled publishing process, not a file-export shortcut. Begin with a trustworthy source, build semantic structure, choose the correct format, preserve accessibility, reconcile metadata, validate the package, and inspect the retailer-processed result.
The strongest workflow catches problems before upload and keeps evidence of what was tested. That produces an eBook that is easier to distribute, update, and read across the environments that matter.
