Passing an Accessibility Checker Isn't the Same as Conforming to PDF/UA-1

A previous post in this series covered the difference between Section 508 and PDF/UA-1. This one covers a distinction inside PDF/UA-1 itself: the difference between passing an automated accessibility check and actually conforming to the standard.

What “the standard” actually means

PDF/UA-1 (ISO 14289-1:2014) is a conformance standard. It doesn't just describe good practice, it defines specific, testable requirements a document must meet. The PDF Association's Matterhorn Protocol translates those requirements into 31 checkpoints covering 136 individual failure conditions. Most of those conditions can be checked by software. A meaningful portion require human judgment, such as whether alt text is actually meaningful, not just present.

What an automated checker can and can't tell you

Tools like Adobe Acrobat's built-in accessibility checker, veraPDF, and PAC (PDF Accessibility Checker) all run automated checks against some portion of these requirements. That's genuinely useful. It catches real problems fast.

But automated checks have known limits, and different tools don't always agree with each other. A PDF that passes one tool's check can still fail conformance against the full standard, because some failure conditions are structurally impossible to verify without a person looking at the document. A missing tag is easy for software to catch. Whether a table's reading order actually makes sense to a screen reader user often isn't.

This isn't a flaw specific to any one tool. It's inherent to the difference between automated pattern-matching and full conformance testing.

Why this matters for AEP document production

An ANOC, EOC, or Summary of Benefits document that shows a green checkmark in one accessibility tool isn't necessarily audit-ready. A carrier that treats “passed the built-in checker” as equivalent to “PDF/UA-1 conformant” is carrying more compliance risk than the checkmark suggests, particularly on the failure conditions that require human review rather than automated scanning.

The fix isn't to distrust every accessibility tool. It's to know which layer of the problem a given tool actually covers, and not assume a pass on one layer means conformance across all of them.