This is the most common format decision in ebook publishing, and the answer is almost always EPUB. But "almost always" isn't "always," and the exceptions matter. The fundamental difference is architectural: EPUB stores structured content that adapts to the display. PDF stores a fixed visual representation that looks identical everywhere. One format was designed for reading on screens of any size. The other was designed for printing on paper.

The confusion persists because PDF is universal. Everyone can open a PDF. Every device renders it. It's the format non-technical people reach for when they want to share a document. But universal support doesn't mean universal suitability. A PDF of a novel on a 6-inch Kindle screen is tiny text on a tiny page. You can zoom in, but then you're scrolling horizontally through every line. That's not reading. That's a UX failure.

How Reflowable Layout Works

EPUB's content is HTML and CSS. When a reading app opens an EPUB, it renders the HTML into pages dynamically, based on the screen size, the user's font size preference, and the user's margin preference. A 1000-word chapter might fill 3 pages on a phone, 2 pages on a tablet, or 1 page on a desktop app. The text reflows to fill the available space.

What the Reader Controls

In a reflowable EPUB, the reader (the person, not the app) controls:

  • Font family — Switch between serif, sans-serif, OpenDyslexic, or any installed typeface
  • Font size — Typically 8 presets from tiny to very large, sometimes a continuous slider
  • Line spacing — Compact, normal, or expanded leading
  • Margins — How much whitespace around the text
  • Justification — Left-aligned or justified
  • Background color — White, sepia, dark mode, black
  • Brightness — Independent of system brightness on some apps

This customization is the entire point of reflowable content. A user with dyslexia can set OpenDyslexic at 18px with wide line spacing. A speed reader can set a dense serif at 10px with narrow margins. A night reader can use white-on-black. None of these adjustments are possible with PDF — you get whatever the author designed.

What Reflow Breaks

Reflow isn't free. When text reflows, spatial relationships between elements are destroyed. A caption next to a figure may end up on a different page from the figure. A table that was designed to fit one printed page may span three screens on a phone. A two-column layout becomes a single column with the left column followed by the right column — or worse, interleaved paragraphs from both columns.

These breakdowns are why certain content types genuinely need fixed layout. The question isn't "is reflow better?" — it's "does this content survive reflow without losing meaning?"

How Fixed Layout Works

PDF describes positioned elements on pages of a fixed size. Each character has an exact (x, y) position. Each image has exact dimensions and placement. Each page has a specific width and height (typically 8.5" x 11" or A4). When you open a PDF, the renderer draws exactly what the author specified. If the screen is smaller than the page, the renderer scales the page down, and everything gets smaller proportionally.

The Small Screen Problem

A standard US letter PDF (8.5" x 11") rendered on a 6" Kindle Paperwhite screen is scaled to 38% of its designed size. Text set at 12pt becomes effectively 4.5pt — below the legibility threshold for most readers. The user can zoom in, but now they see a fragment of the page and must scroll horizontally to read each line. On a phone (5-6.5" screen), the situation is even worse.

PDF readers have a "reflow" mode that extracts text and re-renders it, but this is a lossy heuristic process. Headers may appear in the wrong place. Footnotes may merge into body text. Multi-column layouts often become garbled. Tables frequently break. It's better than nothing, but dramatically worse than a purpose-built reflowable EPUB.

EPUB 3 Fixed Layout: The Hybrid

EPUB 3 supports fixed-layout pages via the rendition:layout pre-paginated metadata property. Each page has a defined viewport (e.g., 1200x1600 pixels) and content is positioned absolutely, like a PDF. The difference is that EPUB3 FXL uses HTML/CSS positioning, so the text is still selectable, searchable, and accessible to screen readers. It can also include some reflowable pages alongside fixed pages in the same book.

EPUB3 fixed layout is the right format for content that needs designed pages but also needs to work within the ebook ecosystem (bookstores, library apps, reading platforms). Children's picture books, cookbooks, and comics increasingly use EPUB3 FXL instead of PDF for digital distribution.

Feature-by-Feature Comparison

FeatureEPUB (Reflowable)PDF (Fixed)EPUB3 FXL
Screen adaptationReflows to any screenScales down (shrinks)Scales down
Font controlUser chooses font/sizeAuthor-fixedAuthor-fixed
Dark modeFull supportWhite page on dark bgPartial
Text-to-speechExcellentUnreliableGood
SearchFull textFull text (usually)Full text
AnnotationsText-level highlightingArea-based markupText-level
Layout fidelityNone (by design)Pixel-perfectPixel-perfect
Image-text spatial relationshipsBroken by reflowPreserved exactlyPreserved
AccessibilityStrong (semantic HTML)Varies (needs tagging)Strong (HTML-based)
File size (typical novel)0.5-3 MB1-5 MB2-10 MB
Phone readabilityExcellentPoorPoor-Fair
Print fidelityLow (depends on renderer)ExactGood

When EPUB Is the Clear Winner

EPUB is the right format for all content that is primarily flowing text:

  • Novels and fiction — Pure text with occasional section breaks. Reflowable by definition
  • Non-fiction prose — Business books, biography, history, self-help. Text with occasional images or pull quotes
  • Poetry — Line breaks are structural (EPUB preserves them), not visual
  • Technical writing with code — Code blocks render in monospace; body text reflows around them
  • Long-form journalism — Articles, essays, collections
  • Religious texts — Bibles, Qurans, etc. Readers need to control font size for extended reading sessions

The common thread: content where the words matter and the visual arrangement is secondary. If you'd be comfortable reading it as plain text (with basic formatting), it belongs in EPUB.

When PDF Is the Right Choice

PDF is correct when the spatial arrangement of elements on the page is part of the content's meaning:

  • Textbooks with figures and sidebars — When a diagram needs to appear next to the paragraph that references it, reflow breaks the relationship
  • Sheet music — Musical notation is a spatial language. Notes, rests, dynamics, and articulations have specific positions on the staff. Reflow is nonsensical
  • Mathematical proofs with equations — Multi-step derivations where the spatial layout shows the progression. EPUB3 supports MathML, but complex layouts still work better fixed
  • Architectural and engineering drawings — Schematics, floor plans, wiring diagrams. Position IS meaning
  • Art books and photography — Full-bleed images, careful typography, designed spreads. The page design is the product
  • Manga and graphic novels — Panel layout, reading flow, and visual storytelling depend on fixed positioning. (Though CBZ is typically better than PDF for this)
  • Academic papers — Two-column layouts with figures, footnotes, and citations. The academic publishing ecosystem revolves around PDF
  • Government and legal documents — Forms, regulations, court filings. Official document format

Converting Between EPUB and PDF

The quality of conversion depends entirely on the direction:

EPUB to PDF: Straightforward

Converting EPUB to PDF is a rendering operation. The converter takes the reflowable HTML content and renders it into fixed pages at a specific size (typically A4 or US letter). Text is laid out, page breaks are inserted, and the result is a multi-page PDF. Quality is generally good — you lose reflowability and user font control, but the content transfers cleanly.

The main decisions are page size, margins, and font. Different converters make different choices, and the results can look quite different. Calibre's EPUB-to-PDF converter lets you configure all of these. ChangeThisFile uses Calibre's defaults, which produce clean, readable output for typical prose books.

PDF to EPUB: Fundamentally Hard

Converting PDF to EPUB is a reconstruction operation, and it's one of the hardest conversions in document processing. PDF stores positioned glyphs (individual characters at specific coordinates), not paragraphs, headings, or semantic structure. The converter must:

  1. Group nearby glyphs into words (detecting spaces by position, not character codes)
  2. Group words into lines (using vertical alignment)
  3. Group lines into paragraphs (using spacing heuristics)
  4. Identify headings (larger font? Bold? Different typeface?)
  5. Detect multi-column layouts and reconstruct reading order
  6. Handle headers, footers, and page numbers (remove or preserve?)
  7. Extract and reposition images
  8. Reconstruct tables (extremely hard from positioned text)

Each step is a heuristic that can fail. Two-column academic papers are the most notorious — the converter often interleaves text from the two columns. Scanned PDFs (images of text) require OCR before any of this can happen. If you have the source document (DOCX, LaTeX, InDesign), always export to EPUB directly. PDF-to-EPUB is a last resort.

Accessibility: EPUB Wins Decisively

For readers who use assistive technology, EPUB is dramatically better than PDF:

  • Screen readers — EPUB's semantic HTML provides heading hierarchy, lists, table structure, and reading order naturally. PDF requires explicit "tagged PDF" structure, which many PDFs lack. An untagged PDF is a bag of positioned glyphs with no reading order — screen readers guess, often incorrectly
  • Text-to-speech — EPUB content is parseable text with correct word boundaries. PDF TTS must reconstruct words from character positions, often stumbling on ligatures, special characters, and column layouts
  • Visual adjustments — EPUB supports user-controlled font size, spacing, and contrast. PDF is fixed — users must rely on OS-level zoom, which doesn't reflow text
  • Braille displays — Work naturally with EPUB's text content. PDF support depends on tagging quality

The EU European Accessibility Act (effective June 2025) and ADA requirements in the US are driving publishers toward EPUB for accessibility compliance. Tagged PDF can meet accessibility requirements, but creating accessible PDFs is harder and more error-prone than creating accessible EPUBs.

The decision is almost always simple: if the content is words meant to be read sequentially, use EPUB. If the content is a designed visual product where spatial relationships matter, use PDF (or EPUB3 fixed layout for ebook distribution). The only complication is content that straddles the line — a textbook with both prose chapters and diagrammatic figures. For those, EPUB3 fixed layout is increasingly the right answer, combining PDF's layout fidelity with EPUB's ecosystem support.

If you have a PDF that should be an EPUB, try the conversion — but set realistic expectations. Simple single-column documents convert acceptably. Complex layouts will need manual cleanup. If you have the source files, skip PDF entirely and export to EPUB from the authoring tool.