Standard
How to check if a document is accessible
Checking whether a document is accessible means testing it against the specific technical criteria that assistive technology relies on, rather than judging it by how it looks. The check has two layers — what a machine can verify, and what only a person can judge — and a trustworthy result reports both honestly.
Start with the machine-checkable criteria. An automated checker can reliably confirm whether the file is tagged at all; whether the structure tree is complete and correctly nested; whether a reading order is defined; whether images have alternative text present; whether the document's language is declared; whether headings, lists, and table headers are marked up; and whether form fields are labelled. These are objective, pass-or-fail conditions, and they catch the majority of common failures — an untagged document, a missing language declaration, images with no alt text at all. Tools that automate this include the PDF Association's Matterhorn Protocol checkpoints (for PDF/UA) and validators built on them, such as the open-source veraPDF, alongside the built-in accessibility checkers in authoring tools.
What automation cannot decide is quality and sense. A machine can confirm alt text exists but not whether it actually describes the image; it can confirm a reading order is defined but not whether that order is the logical one a human would follow; it can confirm headings are marked but not whether the heading hierarchy is meaningful. These require human review. This is why a responsible accessibility check never returns a single "pass" badge for the whole document: it states what was tested, which automated criteria passed, and which items were referred to a person.
A practical workflow is: run an automated check first to find the structural failures and get a baseline; remediate the machine-fixable issues (tagging, reading order, language, artifacts); then have a human confirm the judgement-dependent items (alt-text meaning, reading-order logic, complex tables). Keep a record of what was tested — evidence of the checks that ran is what lets you demonstrate conformance later, rather than merely asserting it. For a single file, the fastest starting point is a free automated check that reports where the document stands against these criteria.
How Normfluss checks for it
Normfluss runs exactly this two-layer check. It validates the file against the machine-checkable PDF/UA and WCAG criteria — tagging, structure tree, reading order, alt-text presence, declared language, table headers, form labels — and automatically remediates the fixable ones. Items that need human judgement, such as whether alt text is meaningful, are routed to a guided review rather than being silently passed. Every check and decision is recorded in a signed evidence record describing exactly what was tested within a stated scope. You can run a free check on a single PDF to see where it stands before committing to remediation.