Pdf ua matterhorn

Real content (text, images, annotations) must be marked as tagged. Artifacts (headers, page numbers, decoration) must be marked as artifact. Our checker flags any untagged content or missing /MarkInfo /Marked true entry.

Custom tag names must map to a standard PDF structure type (/H1, /P, /L, etc.). Tool checks the role map exists and is well-formed; a human confirms the mapping is semantically correct.

Structure elements must be nested correctly — e.g. <LI> only inside <L>, <TH> and <TD> only inside <TR>. Our 161 checks include the full VeraPDF rule set for structural nesting.

All text content must have a Unicode mapping — no private-use glyphs, no font subset with broken ToUnicode CMaps. Screen readers depend on this entirely.

Acrobat's “Suspect” flag in the Accessibility report marks content that looks like it may be missing tags. Matterhorn requires a human to confirm each one isn't a false positive.

Title, language, and the PDF/UA identifier must be present in XMP metadata. Tool checks dc:title, Catalog /Lang, and pdfuaid:part = 1.

Structure tree root, /ViewerPreferences /DisplayDocTitle true, and natural language must all be declared correctly at the catalog level.

Every meaningful image must be tagged as <Figure> with an Alt entry, or marked as an artifact if purely decorative. Tool detects missing alt text automatically; a human decides whether the alt text is meaningful.

Headings must use the <H1>…<H6> structure types and form a logical hierarchy (no jumps from H1 to H4). Tool checks the tags and order; human confirms visual heading style matches.

Real tables must use <Table>, <TR>, <TH>, <TD>. Header cells need a Scope attribute (Row / Column / Both) or an explicit Headers / ID relationship for complex tables.

Lists must use <L> with <LI> children. Each <LI> needs a <Lbl> and <LBody>. Visually “listy” paragraphs that aren't tagged as lists are flagged.

Mathematical expressions must be tagged as <Formula> with meaningful Alt or MathML (PDF/UA-2). Our tool flags formulae that lack alt; a human writes the actual alt text.

Running headers, footers and page numbers must be tagged as artifacts (pagination / header / footer). If they're left as real content, screen readers read them on every page.

Information must not be conveyed by colour alone. Contrast ratios (4.5:1 normal, 3:1 large) are a WCAG concern but widely audited alongside PDF/UA. Tool samples contrast automatically where foreground+background are detectable.

Fonts must be embedded (full or subset) with correct ToUnicode CMaps and Widths arrays. Checker validates every font in the resource dictionary.

Security settings must not block assistive technology. If the PDF is encrypted, the /P bit 10 (extract for accessibility) must be set.

When optional content groups are used, the order must be defined and each group must have a descriptive name. Hidden layers cannot hide essential content.

File attachments must have meaningful descriptions. Our checker flags attachments with missing descriptions; a human writes the text.

If article beads are present, they must respect logical reading order. Tool validates thread structure against reading order tree.

Signature fields must have accessible names and descriptions. The visible signature appearance must be tagged, not left as raw XObject content.

Blank lines, check boxes rendered as images, and similar “print-and-fill” forms must be either converted to interactive form fields or accompanied by an accessible alternative.

Every annotation in the Annots array must either be tagged in the structure tree or marked as an artifact. Stamps, rubber-band highlights, and comments are common offenders.

Every page that contains annotations must declare /Tabs /S (structure order). Visual order, forms logic, and screen-reader order then all agree.

Every form field must have a TU tooltip (screen-reader label) and, where applicable, a T partial name. Required fields must be programmatically indicated.

Link annotations must be inside a <Link> structure element with a /Contents entry or a child text run that describes the purpose. “Click here” is not descriptive.

Audio and video must have captions or transcripts. Our tool flags media annotations; the captions themselves need a human to verify they're accurate and synchronised.

Attached PDFs must themselves conform to PDF/UA. Attached non-PDF files must have accessible descriptions explaining their content.

JavaScript, launch, and navigation actions must not rely on timing or sensory cues that fail for assistive tech users. Flashing content (>3 Hz) is forbidden entirely.

Dynamic XFA forms are not allowed in PDF/UA-1. Static XFA is tolerated only if accompanied by an accessible AcroForm equivalent. Tool flags any dynamic XFA root immediately.

Documents with more than a handful of pages must include bookmarks (outline entries) that reflect the heading structure. Page labels must be human-readable (e.g. “iii”, “3”, “A-3”).

The displayed page number (in headers/footers) and the PDF page label must match. Mismatches confuse screen-reader navigation, cross-references, and citation workflows.