XML, EPUB and PDF Publishing: How to Choose the Right Output Format

Choosing the wrong digital publishing format costs more than a rework cycle — it means content that fails indexing validators, ebooks rejected by retailers, and articles invisible to citation databases. The three core digital publishing formats — XML, EPUB, and PDF — each serve a distinct purpose, and no single format does everything. XML structures content for indexing and multi-format generation. EPUB 3 delivers reflowable, accessible ebooks to readers. PDF preserves a fixed visual layout for print and formal documents. Understanding when each format is the right choice — and when all three are required simultaneously — is the decision this guide is built to support.

The Three Core Digital Publishing Formats at a Glance

XML Publishing: JATS, BITS, and Why XML Must Come First

XML is not a reading format. Readers do not open XML files on their devices. XML is a production format — the structured source from which every other distribution format is derived. This is why the sequence matters: XML must be created first, not last.

JATS XML: What Journals Need and Why

JATS XML is the format that determines whether a journal article appears correctly in PubMed, CrossRef, Scopus, DOAJ, and Web of Science. Every structural element of an article — contributor affiliations, funding statements, reference DOIs, figure captions, supplementary data links — must be correctly tagged in JATS for database systems to ingest, index, and link it accurately. A visually correct PDF alongside an invalid or incomplete JATS file means an article that is present in the journal but missing or incorrectly represented in the databases where researchers actually search for it.

What must be correctly tagged in JATS XML:

• Contributor affiliations — structured with ROR (Research Organization Registry) identifiers, not plain text. Plain-text affiliations cannot be matched to institutional records in Scopus author profiles.
• ORCID identifiers — tagged in <contrib-id contrib-id-type="orcid"> elements for each named author. Missing ORCID tagging affects author disambiguation across citation databases.
• Funding statements — tagged in <funding-group> elements for CrossRef Funder Registry and open access compliance reporting. Funding data present only in the PDF and absent from JATS XML is invisible to funder compliance systems.
• Reference DOIs — each cited reference tagged with a live DOI in <pub-id pub-id-type="doi">. Unlinked references cannot be tracked by CrossRef Cited-by and do not contribute to citation metrics.
• Mathematical notation — encoded in MathML within the XML, not converted to images. MathML is searchable, accessible to screen readers, and correctly rendered by modern platforms. Image-based equations are unsearchable and fail accessibility checks.
• Abstract structure — tagged with correct <abstract> elements including abstract-type attributes for structured abstracts. Untagged abstracts are not correctly extracted by PubMed for display in search results.

BITS XML: What Books and Monographs Need

BITS (Book Interchange Tag Suite) is the book-level equivalent of JATS, designed for structured encoding of monographs, edited collections, and reference works. It is used by institutional library aggregators — ProQuest Ebook Central, EBSCO eBook Collection, and similar platforms — for metadata ingestion, chapter-level indexing, and access management. Publishers supplying content to these platforms without valid BITS XML either cannot be ingested at all or are represented with incomplete chapter and metadata records.

XML-First vs PDF-First: Why the Sequence Is the Decision

The practical consequence: XML-first production eliminates the most expensive class of rework in digital publishing — late-stage XML errors — and allows simultaneous multi-format delivery. Siliconchips Services operates XML-first production as standard across its digital publishing services, with JATS and BITS XML created as primary production files and schema validation reports provided alongside every delivery.

EPUB Publishing: EPUB 3, EPUB 2, Fixed Layout, and When Each Applies

EPUB 3 Reflowable vs Fixed-Layout: Choosing Correctly

EPUB 3 exists in two structural variants. Choosing the wrong one creates a poor reader experience that cannot be corrected by styling alone — it requires rebuilding the file structure.

Reflowable EPUB 3 — text and images reflow to fit any screen size, font size setting, or orientation. The reader controls their preferred reading experience. This is the correct format for all text-heavy content: novels, academic monographs, humanities titles, self-help, business books, and most STM titles where prose and equations are the primary content. The global eBook market is expected to reach $56.51 billion in 2026 and $87.43 billion by 2030, with reflowable EPUB 3 as the dominant distribution format across this growth.

Fixed-layout EPUB 3 — content is positioned at pixel-precise coordinates, maintaining exact layout regardless of screen size. Required for illustrated children’s books, art books, photography titles, and cookbooks where the spatial relationship between text and image is integral to the reading experience. Fixed-layout EPUB 3 does not provide a good reading experience on small screens — the reader must zoom and pan — so it is inappropriate for text-heavy content regardless of how complex the formatting appears in the InDesign source.

EPUB 2 vs EPUB 3: Why EPUB 2 Has No Valid Use Case in 2026

What a Valid EPUB 3 File Must Include

• Passes EPUBCheck 5.x validation with zero errors and zero warnings.
• EPUB Accessibility 1.1 metadata present in content.opf: schema:accessMode, schema:accessibilityFeature, and schema:accessibilitySummary populated correctly.
• Semantic navigation in nav.xhtml with a complete table of contents linked to all chapters and major sections.
• All images have descriptive alt text — not empty attributes (alt="") or filenames (alt="figure1.png").
• Mathematical content encoded in MathML for STM and academic titles — not rendered as images.
• File tested on minimum three environments: Kindle (via Send-to-Kindle), Apple Books, and a desktop EPUB reader such as Adobe Digital Editions.
• Metadata block complete: ISBN, language, publication date, subject classification (BISAC or Thema codes), publisher, and rights statement.

Siliconchips Services provides EPUBCheck 5.x validation reports as a standard deliverable alongside every EPUB file — not available on request. Full-scope EPUB conversion services cover reflowable and fixed-layout EPUB 3, with accessibility metadata and device testing included.

PDF Publishing: Accessible PDF vs Print PDF — and When PDF Is the Right Choice

Accessible PDF vs Untagged PDF: The Structural Difference That Determines Compliance

The majority of PDFs in academic publishing are not accessible — they are visually correct but structurally empty. An untagged PDF has no semantic hierarchy: a screen reader encounters a stream of text with no indication of what is a heading, what is a caption, what is body copy, or what order elements should be read in. For readers using assistive technologies, an untagged PDF of a journal article is effectively unreadable.

The European Accessibility Act, enforceable from June 2025, requires PDFs distributed in EU markets to meet WCAG 2.1 AA standards — which requires tagged structure. Publishers who have not audited their PDF output for accessibility tagging are distributing non-compliant content in their most important institutional markets.

Common Accessible PDF Failures

• Missing document language declaration. The PDF language must be declared in the document properties — not inferred by readers. Missing language declaration fails PAC 2024 validation and prevents screen readers from selecting the correct voice and language rules.
• Untagged figures. Images without tagged alt text in the PDF structure are invisible to screen readers. Every figure, table, chart, and decorative image must be tagged — decorative images with empty alt attributes, content images with descriptive text.
• Incorrect reading order. In multi-column PDF layouts, the logical reading order (left column top to bottom, then right column top to bottom) frequently differs from the visual layout order. Screen readers follow the tag tree order, not the visual order — an incorrectly tagged PDF reads column text in the wrong sequence.
• Equations as images without MathML. Mathematical notation rendered as images in PDF is inaccessible to screen readers. Tagged accessible PDF for STM content should include an ActualText attribute on equation images, or MathML delivered via an associated HTML version.
• Missing or incorrect heading structure. PDFs exported from typesetting software frequently flatten all headings to the same visual style without maintaining H1/H2/H3 hierarchy in the tag tree. A reader using screen reader navigation by heading level encounters a structurally flat document where no headings are distinguishable.

Print PDF vs Accessible PDF: Key Differences

Accessibility Standards: What Each Format Must Meet in 2026

Format Selection Checklist: How to Choose the Right Output for Your Content

For Academic Journal Articles

• Will this article be submitted to PubMed, CrossRef, DOAJ, Scopus, or Web of Science? → JATS XML required.
• Will this article be available on the journal’s website for reader download? → Accessible PDF required.
• Is this journal subject to open access funder mandates (UKRI, NIH, Horizon Europe)? → JATS XML with machine-readable metadata required.
• Is this article being distributed in EU markets? → Accessible PDF meeting WCAG 2.1 AA required (EAA).
• Does this article contain mathematical notation? → MathML in JATS XML required; image-based equations are insufficient.
• Will article be hosted on an online reading platform (HTML)? → HTML5 from JATS XML source required.

For Academic and STM Books

• Will this book be distributed through Amazon KDP, Apple Books, Google Play Books, or Kobo? → EPUB 3 (reflowable) required; EPUB 2 will be rejected.
• Is this an illustrated, art, or design-intensive title where layout cannot reflow? → Fixed-layout EPUB 3 required instead of reflowable.
• Will this book be supplied to ProQuest Ebook Central, EBSCO, or institutional library aggregators? → BITS XML required alongside EPUB 3.
• Is this book being sold in EU markets? → EPUB 3 must meet EPUB Accessibility 1.1 (EAA compliance).
• Will a print edition be produced? → Separate print PDF with CMYK colour profile and printer marks required.
• Does the book contain LaTeX source with complex equations? → XML-first production with MathML encoding required; post-PDF XML conversion will produce image-based equations.

For Trade and Consumer Books

• Is this a text-heavy narrative title (novel, memoir, self-help, business)? → Reflowable EPUB 3 is the correct primary digital format. PDF should not be the sole ebook format.
• Is this being distributed across multiple platforms? → A single EPUB 3 file distributes to Apple Books, Google Play, Kobo, and Nook; Amazon converts EPUB to their internal format on upload.
• Is accessibility required for institutional or library sales? → EPUB Accessibility 1.1 metadata required in content.opf.

Full Format Comparison Table: XML vs EPUB vs PDF

The table below provides a complete side-by-side reference across the decision dimensions that matter most for academic, STM, and book publishers evaluating their digital publishing format requirements.

For publishers working across all three format types, Siliconchips Services provides fully integrated content conversion services covering JATS XML, BITS XML, EPUB 3 (reflowable and fixed-layout), and accessible PDF — with all validation reports provided as standard deliverables alongside every finished file.

Choosing the Right Digital Publishing Format: The Decision in Summary

The format decision for digital publishing is not a single choice — it is a combination determined by content type, distribution channel, and compliance obligations. Academic journal articles need JATS XML for database indexing, accessible PDF for reader download, and HTML5 for online hosting. Academic and STM books need EPUB 3 for retail distribution, BITS XML for library aggregators, and accessible PDF as a companion format. In all cases, an XML-first production workflow — where structured XML is the source from which all formats are simultaneously generated — is the most efficient and compliant approach available in 2026.

If your current production model generates XML after typesetting, delivers formats sequentially, or lacks EPUBCheck and JATS schema validation reports as standard deliverables, you are operating against a production standard that creates avoidable rework costs and compliance risks.