What actually makes a PDF accessible, and why most aren't

A PDF can look perfectly normal — clear headings, readable text, a sensible layout — and still be close to unusable for someone using a screen reader, because visual clarity and structural accessibility are two entirely different properties, and most PDFs are built with only the first one in mind.

What a screen reader actually needs

A sighted reader recognizes a heading because it's bigger and bolder than the surrounding text — that's a purely visual cue. A screen reader has no way to "see" that a line is bigger; it needs the document to explicitly tag that line as a heading in its underlying structure. Without that tag, a screen reader reads a heading exactly the same as a regular paragraph, and the document's entire logical structure — sections, headings, lists, reading order — effectively disappears for anyone not looking at it visually.

The specific things that make the difference

• Tagged structure — headings, paragraphs, and lists marked as what they actually are, not just styled to look that way, so a screen reader can navigate by heading the way a sighted reader scans a page visually.
• Alt text on images — a text description a screen reader can announce in place of an image it can't interpret, so information conveyed visually isn't simply lost.
• A logical reading order — especially important for multi-column layouts, where the visual left-to-right, top-to-bottom order on the page and the order a screen reader should read content in aren't automatically the same thing.
• Real, selectable text rather than a scanned image — a scan with no text layer is invisible to a screen reader entirely, regardless of anything else about the document.

Checking what you actually have

Accessibility Checker scans a PDF for these specific issues — missing tags, images without alt text, an untagged reading order, and text that isn't actually selectable — and reports what's missing, entirely in your browser. It's a diagnostic, not an automatic fixer: for a scanned document, the fix is running OCR PDF first to add a real text layer; for structural tagging, that generally needs to be added back at the point the document is created (in Word, InDesign, or similar), since a PDF that never had structural information can't have it reliably reconstructed after the fact.