Online Tool Store
· 5 min read
How to Check a PDF Is Actually Accessible
Heshan Fernando
Co-founder & COO
A PDF looks fine. Text is selectable, headings are bold and larger, the layout is clean.
A screen reader announces it as one long paragraph with no structure, reads the two-column page straight across so sentences interleave, and skips every image without comment.
Everything that makes the document navigable visually is invisible to anything that is not looking at it.
Tags are the structure
A PDF’s visual appearance and its logical structure are separate. Bold 18-point text looks like a heading; unless it is tagged as one, nothing knows it is.
Tags provide that structure — headings at their levels, paragraphs, lists as lists, tables with header cells identified, figures with alternative text. They are what lets assistive technology navigate by heading, announce a list’s length, or read a table cell with its column header.
An untagged PDF is a picture of a document. The text may be selectable and it has no structure at all, which means no navigation beyond scrolling.
| Element | Visually | Without tags |
|---|---|---|
| Heading | Bold, large | Ordinary text |
| List | Bulleted | Paragraphs starting with symbols |
| Table | Grid | Cells read in sequence |
| Image | A picture | Silently skipped |
Reading order is separate from tags
A document can be tagged and still read in the wrong order, and this is the failure people miss because it requires actually listening.
Assistive technology follows the tag order, not the visual layout. On a two-column page tagged in the order the elements were created rather than the order they are read, a screen reader reads across both columns — producing alternating fragments of two different paragraphs.
The same happens with pull quotes, sidebars and captions placed after the fact. Visually they sit where they belong; in the tag order they can appear anywhere.
Checking reading order means examining the sequence rather than the appearance, and it is the check most often skipped.
Fix the source, not the PDF
You can repair tags in the PDF itself. It works, and it has to be redone every single time the document is regenerated.
The alternative is fixing the source — using real heading styles in Word or InDesign rather than manually formatted text, adding alt text to images there, marking table headers, setting the document language, and using the accessible export option.
That takes about the same time once and every future export is correct. For a document that gets updated quarterly, the difference over a year is substantial.
The exception is a PDF you did not create and cannot regenerate, where patching is the only option.
Document language is one setting
Small and worth calling out because it is trivial and frequently missing.
The document language tells a screen reader which pronunciation rules and voice to use. Without it, the reader guesses — usually from the system locale — and a French document read with English pronunciation is close to unintelligible.
It is one property in the source document and one line in the PDF metadata, and it fixes a serious problem.
Scanned documents need OCR first
A distinction that determines whether anything else is possible.
A scanned page is an image. There is no text in the file, so there is nothing to tag, nothing to read aloud, nothing to select or search. Adding tags to a scanned PDF produces a structured document containing no content.
OCR generates a text layer from the image, which is what everything else depends on. Its accuracy varies with scan quality, and an OCR layer with errors produces a document that reads aloud incorrectly — arguably worse than one that reads nothing, because the errors are invisible to a sighted reviewer looking at the image.
So the sequence for a scanned document is: OCR, check the text layer, then tag. Skipping the check means tagging over text that may not say what the page says.
Common mistakes to avoid
- Assuming selectable text means an accessible document.
- Checking tags exist without checking the reading order.
- Patching the PDF rather than fixing the source, so the next export loses the work.
- Leaving the document language unset.
- Using a scanned image of text with no OCR layer, which has no text at all regardless of tagging.
How to do it with PDF Accessibility Tag Checker
The PDF Accessibility Tag Checker reads the file in your browser.
- Upload the PDF — it is checked locally and never transmitted.
- Fix the document language first; it is one setting with a large effect.
- Work through missing alt text and reading order problems.
- Make the fixes in the source document and re-export.
The PDF/UA standard defines the technical requirements. Other PDF tools are in the tools directory.
Frequently asked questions
What makes a PDF accessible?
Tags describing structure — headings, lists, tables, reading order — plus alt text on images, a set document language, and real text rather than a scanned picture of text.
Why does reading order matter?
Because assistive technology follows the tag order, not the visual layout. A two-column page tagged wrongly is read straight across, interleaving sentences from two paragraphs.
Should I fix the PDF or the source?
The source, then re-export. Patching the PDF works and must be redone every time the document changes; fixing the source makes every future export correct.
Final thought
Set the document language and check the reading order. One takes seconds and the other is the check nobody runs.