Automated and manual checks
The PDF Association's Matterhorn Protocol lists the ways a PDF can fail PDF/UA, the technical standard for accessible PDF, and marks each one as machine-checkable or requiring human judgment. That split is a useful map of what to expect from any checker.
| Software can tell you | A person has to judge |
|---|---|
| Whether the file is tagged at all, and whether any text sits outside the tags | Whether the reading order makes sense |
| Whether images tagged as figures have alternate text | Whether that alternate text describes the image usefully |
| Whether tables have header cells | Whether the right cells are headers |
| Whether heading levels skip (H1 to H3) | Whether the visible section titles are tagged as headings at all |
| Title, language, bookmarks, tab order, form field names, security settings | Color contrast and meaning conveyed by color |
| Pages with images but no text (probably scanned) | Whether recognized (OCR) text is accurate |
For example, a missing description on a Figure tag is machine-checkable (Matterhorn 13-004), but "headings are not tagged" is a human check (14-001), because software cannot know which line of text was meant as a heading.
Free and low-cost tools
- PacketReady runs in the browser, checks a whole folder of PDFs without uploading them, and explains fixes for Word, Google Docs, Acrobat and InDesign. Use it as a fast first pass.
- PAC 2026 from axes4 is a free desktop app for Windows 10 and 11. It covers the machine-checkable PDF/UA and WCAG tests and includes a screen reader preview and structure view.
- axesCheck is a web-based version of PAC; you upload the file to the service.
- veraPDF is an open-source validator for the PDF/A and PDF/UA standards, suited to technical validation.
- Adobe Acrobat Pro includes Check for accessibility and the tools to fix what it finds (Adobe help); it is a paid product.
- NVDA is a free screen reader for Windows. Listening to a document for a minute tells you more about reading order than any report.
Different checkers apply different rules and can disagree, so do not treat one tool's pass as the final word.
A 10-minute manual check
- Title bar. Open the PDF. The window or tab should show a real title, not a file name.
- Copy and paste. Select a paragraph, copy it, and paste it into a plain text editor. Boxes, gibberish or nothing at all means the text is scanned or not mapped to real characters.
- Listen to the first page. With NVDA running, read from the top. Listen for columns read across instead of down, page headers in the middle of sentences, and missing sections.
- Jump by heading. In NVDA's browse mode, press H to move from heading to heading. The headings you hear should match the section titles you see.
- Images. Listen to each map, chart and photo. The description should tell you what someone looking at it would learn. Decorative images should be silent.
- Tables. Move through a data table cell by cell. Each value should be announced with its column header.
- Links. Link text should make sense on its own: "October 6 staff report", not "click here".
- Color. Look for light gray or colored text and for meaning shown only by color. Body text needs a contrast ratio of at least 4.5:1 (3:1 for large text), which a color contrast checker can measure.
- Forms. Tab through fillable fields. Each should announce what to type, in a sensible order.
You do not need to repeat every step for every weekly packet. Check your templates thoroughly once, then spot-check new kinds of documents and anything that failed before.
Reading PacketReady results
Each file gets one of three results:
- Likely fails: at least one critical problem, such as no tags, scanned pages with no text, security settings that block screen readers, or text that does not map to real characters. Fix these first; they block access to the whole document or large parts of it.
- Needs work: no critical problems, but serious ones such as missing alternate text, no document language or title, tables without header cells, unnamed form fields, or no headings in a long document.
- Looks good (automated checks): nothing critical or serious was found. Moderate items and notes may still be listed, and the manual check above still applies.
Every problem shows the evidence (with page numbers where possible), the WCAG 2.1 success criterion, the related Matterhorn failure conditions, and steps for the program you use.
Keep a record
Download the HTML report for each packet and keep it with the meeting records, along with a note of the manual checks you did. If a resident later asks about a document, you can show what was checked and when, and the CSV makes it easy to track progress across many files.
Check your own files
PacketReady checks a whole packet folder in your browser, with nothing uploaded, and explains every fix in plain English.
Related guides: Tagged PDF, explained ยท Scanned PDFs and OCR
Sources
- The Matterhorn Protocol 1.1. PDF Association
- Web Content Accessibility Guidelines (WCAG) 2.1. W3C
- PDF Techniques for WCAG 2.1. W3C
- PAC 2026 download and system requirements. axes4
- axesCheck. axes4
- veraPDF
- Create and verify PDF accessibility (Acrobat Pro). Adobe
- NVDA screen reader. NV Access