A structured-content approach can reduce repeated production work, improve consistency and make publishing workflows easier to extend.
A manuscript is often written once and rebuilt several times. The same title may need a print-ready PDF, a reflowable e-book, a web version, an accessible HTML page, an XML deposit and a partner-specific content feed. When each output is maintained separately, a correction in one place may not reach the others.
XML-first publishing offers a different model: maintain the content in a structured source and generate presentation formats from it. It is a way to mark what content means—such as a heading, author, abstract, reference, figure caption or accessibility description—so systems can process it consistently instead of guessing from page layout.
PDF is useful, but it is not a content model
PDF remains valuable for print proofs, fixed-layout publications and records of how a page appeared. A converter may extract words, but it can struggle to distinguish a heading from bold body text, read a complex table correctly, recover citation relationships or infer the role of an image.
A structured source separates meaning from appearance. One heading element can be rendered with different typography on a website, in a book and in a mobile reading app. A reference can be exported with linked metadata, while an image can carry a caption and alternative text. The publisher can still create a designed PDF; the difference is that the PDF is one output of a reusable content record rather than the master copy from which everything else must be reconstructed.
Start with the content model
Before selecting software, decide which elements matter to the publication. For a scholarly article, that may include title, contributor names and affiliations, abstract, keywords, sections, references, funding statements, figures, data availability and license information. For a technical manual, it may include procedures, warnings, numbered steps, parts lists, illustrations and revision history.
A content model defines each element, its meaning, where it can appear and the rules that make a record valid. Without it, well-formed XML can still carry ambiguous data. Design the model around editorial and distribution needs, not one vendor’s capabilities.
For journal articles, the Journal Article Tag Suite (JATS) is widely used to exchange structured scholarly content.
NISO’s ANSI/NISO Z39.96-2024 (JATS 1.4) standard describes XML elements for article text, graphics and related material. JATS4R recommendations promote consistency and reuse. Choose a profile that fits your content and partners; not every system accepts the same subset. [1][2]
What changes in an XML-first workflow?
A sustainable pipeline starts with editorial content and metadata entering a controlled source. Validation checks the structure and required fields. A transformation layer then produces the formats needed for distribution: web HTML, EPUB, PDF, XML deposits or partner feeds.
Metadata needs to be part of the workflow, not an export task at the end. Contributor names, affiliations, identifiers, publication dates, references, licenses, language, funding and version relationships affect linking, indexing and downstream reuse. Clear responsibility for these fields reduces the number of corrections discovered only after a title is published.
Five benefits that readers and teams can see
First, corrections become easier to propagate. When the source is authoritative, a spelling or reference correction can be applied once and regenerated across outputs. Quality assurance still has to verify each final format, but editors no longer need to locate every separately maintained copy.
Second, outputs become more consistent: shared rules standardize headings, references, tables and metadata. Third, new channels are easier to support; a partner XML profile or API feed may need a mapping rather than a manual rebuild.
Fourth, structure supports accessibility. Semantic headings, reading order, table relationships, image descriptions and language metadata are easier to check when captured deliberately. W3C’s EPUB Accessibility 1.1 specifies accessibility and discoverability requirements for EPUB publications, including accessibility metadata. Structure alone does not guarantee conformance, but it provides a better basis for testing. [3]
Fifth, richer metadata improves discoverability and system-to-system exchange. Crossref’s metadata guidance emphasizes accuracy and completeness for scholarly records. Consistent identifiers, contributor data and reference links help content move between publishers, repositories, libraries and indexing services. [4]
Practical example: one article, five outputs
Imagine a scholarly publisher releases an article as a journal web page, a print PDF, a full-text XML record, an EPUB collection and a repository deposit. This is an illustrative scenario. In a file-by-file process, a late correction to an author affiliation might be made on the web page but missed in the repository deposit or XML archive.
With a structured source, the correction is made in the article record and the affected outputs are regenerated. Automated validation checks required metadata and link structure, while an editor reviews the rendered web page, PDF and e-book.
The publisher can measure whether fewer mismatches reach production and how long corrections take. Each format can keep its own design while representing the same approved content; the goal is consistency of the source, not identical presentation everywhere.
Migrate without stopping production
Most organizations cannot convert a full catalog overnight. Start with a bounded slice: a new journal, a high-volume title type or a workflow with frequent reformatting. Inventory source files, identify recurring defects and map the minimum metadata needed by each output. Select a standard or profile, define governance, then run parallel production long enough to compare accuracy and cost.
Legacy conversion should include exception handling. Scanned documents, complex tables, equations and older files may need editorial correction after conversion. Do not treat “XML created” as proof of success; validate the content and review representative outputs. Keep original files and a traceable relationship to the converted record so the migration remains auditable.
Measure the workflow, not just the files
Useful measures include production time per title, manual rework hours, correction propagation time, validation failures, percentage of outputs passing quality checks and metadata completeness. Track the cost of maintaining the content model and transformations too. If the model is too complicated for editors to use, the process may simply shift effort from formatting to data entry.
XML-first publishing works best when people agree on the source of truth, standards, ownership and exception process. Begin with a content model tied to genuine needs, automate repeatable transformations and keep human review where interpretation and presentation matter. The goal is not “XML everywhere” for its own sake; it is to make content easier to maintain, verify and reuse.
Sources cited and further reading
1. NISO, “ANSI/NISO Z39.96-2024, JATS: Journal Article Tag Suite, version 1.4” ↩
2. JATS4R, recommendations for structured, reusable scholarly content ↩