What an Automated PDF Accessibility Scan Can and Cannot Detect

Document title present or missing. A scan checks whether the PDF has a title in its metadata. Many PDFs ship with the filename as the title – or no title at all. Document language set or not set. Screen readers use the declared language for pronunciation rules. A scan verifies whether the document specifies its language.

Tag presence or absence. A scan detects whether a PDF has been tagged at all. An untagged PDF is essentially an image stack to a screen reader. Tag type correctness. Is a heading tagged as a heading? Is a paragraph tagged as a paragraph? Is a table tagged as a table? Scanners verify these tag assignments against the actual content. Heading hierarchy issues. Skipped heading levels (jumping from H2 to H4) and missing H1 tags are easy to detect programmatically.

Image alt text presence or absence. A scan can check whether every image has an associated alt text entry. This is one of the most common automated checks. Table header associations. Scanners can verify whether table cells are associated with header rows using proper scope or ID associations. List structure correctness. Is the list tagged as a list? Are the list items properly nested? Automation checks structural list markup. Form field labels. A scan verifies whether form fields have associated labels in the tag tree. Character encoding issues. Scanners detect characters that are not properly encoded, which can cause screen readers to mispronounce or skip text. These checks are objective. There is no ambiguity about whether a document has a title or an image has alt text. The answer is yes or no, and a scanner gets it right.

A scan can tell you that an image has alt text. It cannot tell you that the alt text says "image_01_final.jpg" instead of something useful. The programmatic check passes. The accessibility check fails. Only a human reading the alt text alongside the image can judge its quality.

Scanners calculate contrast ratios against WCAG thresholds. But a document might pass every ratio while still using color as the only way to distinguish categories in a chart. The numbers look fine. The document is still hard to use for someone who is colorblind.

A scan can verify that the tag tree has a logical order and that content flows top to bottom without jumps. But it cannot tell whether that order matches how a human would actually read the page. A sidebar might come before the main article in the tag sequence, which is technically ordered but practically confusing.

A scan can confirm that a table has headers with proper cell associations. It cannot tell whether a twelve-column table with merged cells and nested headers is comprehensible when read aloud. The structure is valid. The user experience may not be.

Scanners can verify that a link has alternate text and that the destination is not broken. They cannot tell whether the link text makes sense out of context or whether the destination page is itself accessible. "Click here" passes the link check. It fails the usability check. This middle zone trips up many teams. They see a scan report with few issues and assume the document is in good shape. The real question is what kind of issues remain – and only human review can answer that.

A person must decide whether alt text describes the image well enough that someone who cannot see it still understands what it conveys. A scan cannot judge whether "Bar chart showing 23% increase in Q3" is better than "Chart."

Only a human can read through a document with a screen reader and determine whether the sequence makes sense. Does the caption come after the image? Does the footnote interrupt the main text? These are experiential questions, not structural ones.

Scanners have no ability to evaluate whether a complex flowchart or data visualization is adequately described. A long paragraph of alt text might technically exist, but a human needs to judge whether that description actually communicates what the diagram shows.

A scan might flag that color is used in a chart. But a human must determine whether the same information is available through pattern, label, or text. Two bars might be red and green – the scanner sees colors. A person must decide whether the values are also labeled directly.

Does the document make sense when read from beginning to end? Are sections organized logically? Is the writing clear? These are questions about content quality that no scanner can answer.

A scan can verify that form fields have labels. It cannot tell whether the tab order is logical, whether error messages are clear, or whether the form is actually usable with keyboard navigation alone. Only testing with assistive technology reveals these issues.

Ultimately, accessibility is about whether people with disabilities can use your documents. A scan cannot simulate that experience. It cannot tell you whether a screen reader user would find your annual report useful or frustrating. That requires user testing or expert manual review.

Initial triage. Run every PDF through a scan first. It takes minutes and surfaces obvious structural problems immediately. Bulk assessment. If you manage hundreds or thousands of PDFs, scanning gives you a dataset. You can see which documents have the most issues and where to focus remediation effort. Issue categorization. Scan reports group problems by type – missing tags, heading issues, alt text gaps. This helps you plan remediation work in batches. Progress tracking. Rescanning the same PDF after remediation shows whether structural fixes took effect. It is a useful checkpoint in your workflow.

Samples from scan results. After scanning a batch, pick representative documents for manual review. Focus on high-visibility documents, complex layouts, and documents flagged for borderline issues. Complex documents. Annual reports, technical manuals, forms, and documents with charts or diagrams almost always need human eyes. Final quality assurance. Before publishing or submitting documents for compliance review, have someone test them with a screen reader or send them to an accessibility expert.

Scan first. Categorize the issues. Prioritize by document importance and user impact. Remediate the structural problems. Then spot-check a sample with manual review. Rescan to confirm the fixes. Repeat. This approach saves time because you are not manually reviewing every PDF from scratch. It also catches the interpretive issues that automation misses. The scan is a filter, not a finish line.

No. An automated scan checks structural and programmatic criteria, but many WCAG success criteria require human interpretation. A scan can identify missing properties and structural problems. It cannot judge whether content is meaningful, whether reading order is logical, or whether complex visuals are adequately described. Use scans as part of a broader workflow that includes manual review.

Most accessibility professionals agree that automated tools can detect roughly 25 to 30 percent of WCAG success criteria reliably. Another 20 to 30 percent can be partially flagged but require human verification. The remaining 40 to 50 percent are interpretive criteria that automation cannot evaluate at all. The exact split depends on document type and content complexity.

Yes, for documents that matter. A scanner is excellent for bulk assessment and structural issue detection. For public-facing documents, legal submissions, and complex layouts, human review is still necessary. Many organizations use scans to narrow the scope of expert review – which makes the expert's time more efficient and less expensive.

Scan new PDFs before they are published. Scan existing PDFs on a regular schedule – quarterly or bi-annually depending on volume. Rescan after remediation work to verify fixes. If your organization produces PDFs continuously, integrate scanning into your production workflow rather than treating it as a one-time check.

Not necessarily. A clean scan means the document passed the checks that automation can perform. It may still have poor alt text, confusing reading order, or usability problems that only human testing reveals. A clean scan is a good sign. It is not a final verdict.