Status: researched on 2026-08-24. Current RusDox output is not tagged PDF, not PDF/UA, and not PDF/A. PdfRenderEvidence records these claims as false. A language entry, embedded fonts, Unicode maps, and source-level image descriptions are useful prerequisites, but they are not conformance.
Tagged PDF and PDF/UA
The current accessibility standard is ISO 14289-2:2024 (PDF/UA-2). The PDF Association sponsored-standards programme provides free access to the relevant ISO standards, and its Tagged PDF Best Practice Guide explains the syntax implementation boundary.
An honest first tagged-PDF milestone needs all of the following before any accessibility claim:
/MarkInfowith marked content and a/StructTreeRootrooted in a semantic document element;- marked-content IDs, structure elements, and a correct parent tree that bind every meaningful page object to the logical structure;
- semantic roles for headings, paragraphs, lists, tables, links, notes, and figures, including correct table headers and nesting;
- logical reading order independent of paint order, with decorative content marked as artifacts;
- alternative text for figures, replacement or actual text where needed, a catalog language, and language changes within mixed-language content;
- embedded fonts with reliable Unicode mapping, tagged annotations, document title, viewer preferences, and the required XMP identification metadata.
The W3C PDF techniques provide focused implementation checks for reading order, image alternatives, document language, language changes, and artifacts. The Matterhorn Protocol is a useful test catalogue, not a substitute for the standard.
CI should run the exact PDF/UA flavour of veraPDF validation and retain the machine-readable report. A veraPDF pass is necessary but not sufficient: release evidence must also include manual reading-order inspection and tests in representative assistive technology. RusDox will not infer conformance from a generic PDF validity check.
PDF/A archival output
The PDF/A overview separates archival profiles by PDF generation and semantic level. The recommended first RusDox target is PDF/A-2u: it fits the current PDF generation more closely and requires reliable Unicode mapping without prematurely claiming a complete logical structure. PDF/A-2a plus PDF/UA-1 can follow after tagged structure is implemented. PDF/A-4 remains a later evaluation because the ISO 19005-4:2020 page currently marks that edition as under revision.
PDF/A-2u implementation must, at minimum:
- embed every font and provide Unicode mapping for all text;
- embed a valid ICC output intent and make colour use self-contained;
- add correct PDF/A XMP identification and any required extension schemas;
- prohibit encryption, unresolved external rendering dependencies, and actions or JavaScript forbidden by the selected profile;
- define policy for attachments and annotations under the exact chosen part and conformance level;
- pass the explicit PDF/A-2u flavour in veraPDF, not merely auto-detection.
The PDF Association publishes guidance on XMP identification and extension-schema templates. The open veraPDF validation profiles make the automated rule set reviewable.
Claim gate
RusDox may advertise tagged PDF, PDF/UA, or PDF/A only after the renderer emits the complete required structures and metadata, fixtures exercise every relevant feature, the exact veraPDF flavour passes in CI, and the human checks required for accessibility are recorded. Until then, API evidence remains false and the public compatibility matrix says unsupported rather than “nearly compliant.”