Standards

Two standards, one validation pipeline.

What WCAG 2.2 covers, what PDF/UA-1 covers, and why we validate every PDF against both before it ships.

WCAG 2.2 (full) and PDF/UA-1 ride together

WCAG 2.2 is the W3C's 2.2-level Web Content Accessibility Guidelines applied to the PDF surface — covering user-facing experience and the rules a sighted-user accessibility audit cares about. PDF/UA-1 (ISO 14289-1:2014) is the PDF Association's standard for tagged accessible PDFs — covering tag-tree machine-readability and the rules assistive technology actually consumes.

A document can pass one and fail the other. WCAG passes when the visible page is OK; PDF/UA passes when the underlying tag tree is OK. Procurement contracts ask for WCAG. Screen readers use PDF/UA. We validate against both so the procurement contract and the end user are both satisfied by the same artifact.

What a tagged PDF actually is

A tagged PDF carries a structure tree alongside its visual content. Every paragraph, heading, list, table, and link has a role assignment that screen readers, voice control, and refreshable braille consume. An untagged PDF that "looks fine in Acrobat" presents to assistive technology as a flat image — visible content, zero semantic structure, unusable.

Three layers are required for assistive tech to navigate the document. The structure tree maps each on-page mark to a logical role (H1, P, LI, TD, etc.). The role map binds the document's custom tags to standard ones. Marked content connects the tree back to the rendered glyphs. Missing any one of the three breaks the chain.

What the failures look like

Four rule families dominate every corpus we have run. Each maps to a specific user impact — not an abstract compliance number.

7.1:3

Content untagged

Document has no structure tree. Screen readers see a flat image; nothing is reachable.

7.18.1:1

Document title missing

Tab labels render as "Untitled". The user has no anchor for which document they are in.

7.18.3:1

Natural language not declared

Synthesizer picks the wrong voice and pronunciation. English document read with Spanish phonemes is unintelligible.

7.21.*

Font / ToUnicode broken

Glyphs do not map back to characters. Copy-paste produces random Unicode. Screen readers spell out glyph indices.

Why veraPDF is authoritative

veraPDF is built by the PDF Association — the same body that authored PDF/UA-1 — and is the reference validator the standard publishes itself against. Validating against an internal scorecard is self-verification. Validating against veraPDF is third-party verification by the rule-makers.

The site never claims a result veraPDF does not confirm. Every per-corpus number on the proof page was generated by running veraPDF 1.30.0 / 1.30.1 against the PDF_UA/PDFUA-1.xml and WCAG-2-2-Complete.xml profiles.

Matterhorn Protocol

Matterhorn is the PDF Association's checklist of 31 checkpoints (97 sub-checks) implementing PDF/UA-1 in practice. Some are deterministic and machine-checkable — for example, every marked content sequence must be reachable from the structure tree. Some require human judgment — for example, is the alt text accurate to the image, not just present.

The pipeline handles every deterministic checkpoint programmatically and surfaces the human-judgment checkpoints for the customer's accessibility coordinator to review before sign-off. The same per-file rule diff that lives on the proof page is the artifact a coordinator works against.

How the pipeline implements those checkpoints — seven stages, the seventeen verified fixes discovered through veraPDF validation on real court PDFs, and the nine disclosed limitations — is on the engine page. Full rule references are at /engine/pdf-ua-1-rules and /engine/wcag-2.2-criteria.

Buyer Q&A

Do I need both standards?
Most procurement contracts and 508-compliance audits ask for WCAG. PDF/UA is what assistive technology actually reads. Pass both.
What does my legal counsel need?
An auditable per-file pass / fail report tied to a published validator. veraPDF is that.
What does my coordinator verify before sign-off?
Three things: (a) the structure tree exists and has the document's actual headings, lists, and tables; (b) natural language is declared at the catalog level; (c) sample alt text is accurate to the image.

More on regulatory deadlines, scope, pricing, and how automated remediation compares to manual: see the FAQ.