Common PAC and veraPDF errors and how to fix them
Thousands of errors in PAC for a short flyer? There is usually a single cause behind them. PAC and veraPDF report against the Matterhorn Protocol, and their messages are precise, just rarely easy to read.
Last updated:
How the messages relate
The Matterhorn Protocol breaks PDF/UA-1 down into checkpoints covering metadata, fonts, headings, tables and more. PAC groups its results by these areas, veraPDF cites the relevant rule of the standard. Many individual messages share one root. If the structure tree is missing, the validator reports every text object separately. So fix the fundamental errors first, then check again.
For each error below we note whether the EasyPDFUA report shows it and whether retroactive tagging fixes it. Tagging only works on PDFs that have no tags yet.
Metadata and document settings
Document title missing or not displayed
The XMP metadata lacks dc:title, or DisplayDocTitle is not set to true. The screen reader then announces the file name, something like “final_v3_print.pdf”. The fix is to enter a title in the document properties and enable “Show document title”. When tagging, EasyPDFUA writes the title to XMP and DocInfo and switches on DisplayDocTitle.
Default language missing
The document catalogue has no /Lang entry. The screen reader falls back to its own default, and an English text read with a German voice sounds accordingly odd. Set the main language, for example en-GB or de-AT, and give passages in other languages their own language attribute on the relevant tag.
EasyPDFUA sets the default language when tagging. Language changes within the text are yours to mark up.
PDF/UA identifier missing
Without pdfuaid:part in the XMP metadata, a validator does not treat the file as a PDF/UA document. The identifier is added on export through the PDF/UA option or afterwards, ideally only once the other requirements are met. EasyPDFUA writes pdfuaid:part = 1 when tagging. That declares the target profile. It is not a passed validation.
Tagged content and artifacts
Document is not tagged
No StructTreeRoot, no MarkInfo. The usual origin is “Print to PDF” or an export with the accessibility option switched off. Re-exporting the source file with tagging enabled is the best fix.
If the source is gone, EasyPDFUA builds a basic structure tree: one generic P tag with marked content per page. That clears the fundamental errors. It is not yet a meaningful outline of the content.
Content neither tagged nor marked as artifact
Individual text, image or path objects are not attached to any tag. Stamps, headers or graphics added later are frequent culprits. Content that carries information belongs in the structure, purely decorative elements become artifacts. For an untagged PDF, EasyPDFUA assigns the page content to the base structure. Deciding what is content and what is decoration remains your job.
Artifact inside tagged content or vice versa
Artifact and content markings are nested inside each other, almost always after faulty post-processing. The content panel in Acrobat Pro or axesPDF is where this gets untangled. EasyPDFUA stays out of it, because the tagger rejects PDFs that are already tagged.
Fonts
Font not embedded
The PDF refers to a font without supplying its font program, so the rendering depends on the reader’s machine. The clean fix is a new export with embedded fonts, usually a setting in the PDF export options.
EasyPDFUA checks active TrueType fonts for missing font programs and can embed one as /FontFile2 when tagging, but only under strict conditions. A local font file with exactly the same PostScript name must be available, and its OS/2 fsType licence bits must permit embedding. Every character code actually used must be resolvable, and the font’s widths must match the widths stored in the PDF exactly. EasyPDFUA never substitutes a font that merely looks similar. CFF fonts and non-embedded fonts without a provable local counterpart are left unchanged.
Missing Unicode mapping (ToUnicode)
A font has no mapping from character codes to Unicode. The screen reader reads out garbage, and search and copy fail. Re-exporting with a complete font or adding the mapping fixes it.
EasyPDFUA repairs the mapping when it can be derived unambiguously from the embedded TrueType tables, or when an exact glyph match against a locally installed reference font of the same family succeeds. Existing ToUnicode entries are kept. Where glyphs are ambiguous, nothing is changed.
Links, annotations and tab order
Link annotation not inside a Link tag
A clickable area is not bound to the structure through a Link tag with an object reference, or it lacks a description (Contents or alternative text). Every link annotation needs both.
EasyPDFUA does not modify existing link annotations. Where a web or email address stands alone on a line in an untagged PDF, tagging creates a new /Link annotation for it with a Link tag, an OBJR and a ParentTree entry. Addresses inside longer lines are not linked.
Tab order does not follow the structure
Pages with annotations lack a /Tabs /S entry, so the Tab key moves through links and form fields in some other order. In Acrobat the setting is called “Use Document Structure”. EasyPDFUA sets /Tabs /S on every page where it has created links.
Semantic structure
Figure without alternative text
A Figure tag has no Alt attribute. Add text that describes the purpose of the image, or mark decorative images as artifacts. The EasyPDFUA report points out such figures. It does not write alternative text, since that requires understanding the content.
Heading level skipped
An H1 is followed directly by an H3, or headings are merely bold and not tagged at all. The fix lies in the source file: heading styles and a hierarchy without gaps. EasyPDFUA does not create heading tags.
Table without header cells
The table consists only of TD cells, or the TH cells have no Scope. A screen reader then cannot relate a value to its column. Mark header rows as such (in Word, for example, with “Repeat as header row”) and define the scope. The EasyPDFUA report flags tables without TH. The tool does not create table or list tags.
In what order to work
Start by examining the PDF with the PDF/UA check. If it is untagged, let the application add title, language, identifier, base tags and links, then run the result through PAC or veraPDF again. The number of messages usually drops sharply, leaving headings, tables, lists and alternative text. Those you handle by hand or by exporting the source again.
Frequently asked questions
Why does PAC report thousands of errors?
Usually the basic structure is missing. In an untagged PDF every text object counts as a separate error. Once a structure tree exists, the list shrinks considerably.
What is the difference between PAC and veraPDF?
PAC is a free Windows application with a graphical interface and a screen reader preview. veraPDF is open source and at its best on the command line and in automated workflows. Both check the machine-verifiable criteria of PDF/UA.
Can EasyPDFUA fix all PAC errors?
No. It fixes technically unambiguous errors: title, language, PDF/UA identifier, base tags, links for standalone addresses including tab order, and Unicode mappings that can be repaired unambiguously. A missing TrueType font program is embedded only if a local file with the same PostScript name, permitted embedding and exactly matching widths is available. Headings, tables, lists and alternative text remain manual work.
Does “no errors in PAC” mean the PDF is accessible?
Not necessarily. PAC only checks what a machine can decide. Whether alternative text says anything useful, headings are right and the reading order is logical is for a person to judge.