What is PDF/UA? The standard behind accessible PDFs, explained

PDF/UA (Universal Accessibility) is the international standard — ISO
14289 — that defines what a technically accessible PDF looks like on the
inside: complete tags, a machine-readable structure, correct metadata, and
none of the constructs that break assistive technology. Where WCAG describes
accessible content in general, PDF/UA pins down the file format
requirements specifically for PDF. It matters to your district because it's
the standard remediation work is measured against — and because the free
validator that checks it, veraPDF, is how you can verify a vendor's work,
including ours. Here's the plain-language version of both.

PDF/UA in plain language

A PDF can look perfect on screen and still be a locked box to a screen
reader. PDF/UA-1 (ISO 14289-1) exists to make the difference testable. In
substance, a conforming file must:

• carry a complete tag structure — every piece of real content marked as
  what it is (heading, paragraph, list, table, figure), in a logical order;
• mark everything that isn't content (decorations, page furniture) as an
  artifact, so assistive technology skips it;
• provide alternative text for meaningful images, a declared document
  language, and a real document title shown instead of the filename;
• tag tables so header cells are associated with their rows and columns, and
  represent lists, footnotes, and links with the correct structures;
• embed fonts and map every character to Unicode, so text can actually be
  extracted and spoken;
• avoid constructs that defeat assistive technology (like conveying
  information through visual position or color alone).

If that list sounds familiar, it should — it's the formal version of the
checklist in our how-to guide.
PDF/UA is that checklist turned into an engineering specification precise
enough that software can verify most of it.

Two housekeeping notes. First, PDF/UA-1 dates to 2012 and targets the PDF
1.7 format nearly every government document uses today; a successor,
PDF/UA-2, arrived in 2024 for the newer PDF 2.0 format. PDF/UA-1 remains the
practical target for remediation work in 2026. Second, PDF/UA and WCAG
overlap heavily but aren't identical: the DOJ's Title II
rule legally requires
WCAG 2.1 AA, and WCAG adds content rules — minimum color contrast, for
one — that PDF/UA doesn't police. In practice, a document remediated to
PDF/UA with sensible visual design satisfies what WCAG asks of PDFs; PDF/UA
is how the document side of WCAG compliance actually gets built and
verified.

The Matterhorn Protocol: a catalog of every way to fail

A standard is only useful if you can test against it. The PDF Association
publishes the Matterhorn
Protocol — the
industry's agreed testing model for PDF/UA-1. It organizes conformance into
31 checkpoints containing 136 failure conditions: a complete catalog
of every distinct way a PDF can violate the standard, from "table header
cells not tagged" to "document language missing."

The most useful thing Matterhorn does is sort those 136 conditions by who
can detect them:

• 87 are machine-checkable — software can find them with certainty. Is
  there a tag tree? Is the language set? Are fonts embedded? Do table cells
  reference headers?
• 47 require human judgment. Software can see that alt text exists; it
  can't tell whether "image123.png" is a meaningful description. It can see a
  reading order; it can't always know whether that order makes sense.
• 2 aren't cleanly either.

That roughly two-thirds / one-third split is the single most important fact
in this post, and we'll come back to it.

What veraPDF is — and why it's the referee

veraPDF is the open-source validator the
industry treats as the neutral referee for PDF standards. It was built by a
consortium led by the Open Preservation Foundation and the PDF Association,
with funding from the European Union's PREFORMA digital-preservation
project — which is to say, it belongs to no vendor, costs nothing, and its
rules are public. It validates PDF/A (the archival standard) and performs
the machine checks for PDF/UA — the checkable conditions from the
Matterhorn catalog above.

Run veraPDF against a document and you get a precise, itemized verdict:
every machine-checkable failure condition the file violates, by name. That
precision is what makes it valuable to a district in two directions at once:

• As a buyer, you can verify any vendor's output yourself — including
  ours. "Remediated" is a claim; a clean veraPDF report is evidence. A
  vendor whose documents fail veraPDF is selling you something less than
  they've promised, and checking takes minutes.
• As a DIYer, a failing report is a to-do list. Each named failure
  condition tells you exactly what to fix, which beats guessing at what a
  screen reader will trip over.

Why passing veraPDF isn't the finish line

Here's the honest part, and it cuts against anyone who waves a validation
report as the whole story: veraPDF verifies structure, not meaning.
Remember the split — 87 of Matterhorn's 136 failure conditions are
machine-checkable, and those are what veraPDF tests. The other 47 need a
human, and they're disproportionately the ones your patrons actually
experience:

• Alt text that exists but says nothing ("photo," "chart") passes the
  machine check and fails the reader.
• A tag tree that's present and well-formed but tags a heading as a
  paragraph reads as valid structure and broken navigation.
• A reading order that's technically declared can still be the wrong order
  for a two-column newsletter.

So the true quality bar is a conjunction: pass the machine checks and
survive human review of the judgment calls. A document that fails veraPDF
is not accessible; a document that merely passes it might not be either.
Anyone who tells you a green checkmark from a validator equals accessibility
is skipping the 47 conditions software can't see.

What this means for your district

You don't need to memorize checkpoint numbers. You need three takeaways:

1. Ask vendors what standard they remediate to and how they verify it.
   The right answer names PDF/UA and includes machine validation — and
   acknowledges the human-judgment layer rather than pretending a validator
   covers it.
2. Spot-check with veraPDF yourself. It's free, and it turns "trust us"
   into a report you can read. Your April 2028
   deadline makes
   verified work, not just claimed work, the thing worth paying for.
3. Budget human eyes for the judgment calls — especially alt text on
   charts and maps, and reading order on complex layouts.

This split is exactly how Ketchly is built: every document
we remediate is validated with veraPDF against PDF/UA, and the
human-judgment conditions — the ones no validator can score — get review
where the AI isn't confident. The standard is public, the validator is free,
and our work is checkable by anyone; we think that's how it should be.