Accessibility: tagged PDF and PDF/UA

var options = new PdfOptions
{
    Accessibility = PdfAccessibility.PdfUA1,   // or PdfUA2, or Tagged: the same structure without the PDF/UA claim
    Language = "da",                           // default: <html lang="...">; for PDF/UA without either, "en"
    Conformance = PdfConformance.PdfA2b,       // optional: PDF/UA-1 combines with PDF/A-2 and 3, PDF/UA-2 with PDF/A-4
};

Tagged PDF

Tagged gives the PDF a structure tree built from the HTML, so screen readers, reflow and text extraction follow the document rather than the page layout: headings (H1 to H6), paragraphs, lists with their labels, tables with header scopes and spans, figures with their alt text, links (tied to their clickable areas), block quotes, code and sections, all in document order, with lang attributes as the language of the element. Text outside a paragraph (straight in a div, say) gets one. Backgrounds, borders, running headers and footers, HeaderText/FooterText, watermarks, repeated table headers, fixed boxes and decorative images are artifacts that assistive technology skips.

PDF/UA-1

PdfUA1 adds what ISO 14289-1 asks beyond the tags: every font embedded (the bundled Liberation fonts stand in for the standard ones), the title shown in the viewer's title bar and the PDF/UA identification in the XMP metadata. The output passes veraPDF's PDF/UA-1 checks, on its own and together with PDF/A-2b or PDF/A-3b (e-invoices included). It needs a title: Title, the <title> element or else the first heading; without one, conversion throws ArgumentException, as it does when Security leaves out PdfPermissions.ExtractForAccessibility.

PDF/UA-2

PdfUA2 (ISO 14289-2:2024) is the PDF 2.0 successor: the file is written as PDF 2.0, its tags are in the PDF 2.0 structure namespace (block quotes, quotes and code, which PDF 2.0 dropped, keep their PDF 1.7 type and are role mapped to Div and Span), links and bookmarks go to structure destinations (the element they point at, not only its page), lists state their numbering and form fields describe themselves, and the XMP says pdfuaid:part 2 with pdfuaid:rev 2024. It has the same title and language requirements as PdfUA1 and passes veraPDF's PDF/UA-2 checks, on its own and with PdfA4. PDF/UA-1 and PDF/A-2 and 3 are based on PDF 1.7, PDF/UA-2 and PDF/A-4 on PDF 2.0, so PdfUA2 with PdfA2b (or any PDF/A-2 or 3 level) and PdfUA1 with PdfA4 throw ArgumentException.

What your HTML needs

What the tags can't fix is up to the HTML: give images an alt (an image without alt, aria-label or title, or an <svg> without a <title>, counts as decorative), use real headings, lists and th cells, and mark purely visual content aria-hidden="true". Heading levels that skip one (h1 then h3) are tagged one level below the previous heading, as PDF/UA does not allow gaps.