PDF accessibility basics for teams that share documents

A PDF that looks fine can still be a pile of coordinates. Screen readers need tags, a logical order, and text you can select. A desktop export from Word or LibreOffice can keep that structure when tagging is turned on. The browser Word to PDF tool on this site extracts text and drops the tags. Merge PDF and Compress PDF usually do not rebuild tags, so a merged or compressed file needs another check before it goes to a client.

One onboarding PDF looked normal on screen. A screen reader jumped from the footer to a heading to the page number. The file had no reading order, only positions.

The same file has to work for voice navigation, for reflow on a small screen, and for someone searching a clause late at night.

Who runs into the broken file

• Blind and low-vision people using a screen reader
• Dyslexic readers using text-to-speech
• People reflowing the page on a phone
• Search inside a document library
• Anyone copying a quote instead of retyping it

A scan-only PDF is a barrier for all of those.

What the file needs

1. Real text

Exports from Word or Google Docs usually contain text objects. A scan needs OCR, optical character recognition, before there is a searchable layer.

Try to select a sentence. If you cannot select it, text tools and screen readers have nothing to read.

2. Tags

Tags name the structure:

• Headings (H1, H2, and so on)
• Paragraphs
• Lists
• Tables with row and column scope
• Figures with alt text

An untagged PDF is a set of "draw this text here" instructions. It can look right and still be chaos for assistive tech.

3. A logical reading order

Columns, sidebars, and footnotes need an explicit order. Viewers guess, and they guess wrong often. Annual reports are a common case.

4. Alt text on images that mean something

Decorative images can be marked as artifacts and ignored. Charts and diagrams need a text description. An empty alt is not a description.

5. Color and contrast

Light gray body text fails WCAG contrast. Low-vision readers need type they can see, and a path a screen reader can follow.

Export it correctly the first time

Repair is harder.

From Word:

• Use heading styles (Heading 1, Heading 2), not a larger bold font applied by hand.
• Add alt text on inserted images (right-click, then Edit Alt Text).
• Turn on "Document structure tags for accessibility" in the PDF export. On Windows Word that is a checkbox.
• Keep the main text out of floating text boxes.

See Word to PDF for what the browser extract keeps, which is the text, and export from Word when the tags have to survive.

From slides:

• Put titles in the title placeholder, not in a drawn text box.
• Speaker notes do not replace alt text on a chart.

From scans:

• Run OCR in the scan step when you can.
• Do not compress so hard that the OCR text turns into garbage.

Fixing a PDF you already have

1. Run a checker (Acrobat Pro, PAC, or axesPDF, depending on budget).
2. Add tags by hand, or auto-tag and then review. Auto-tag always needs a person to pass over it.
3. Fix the reading-order panel. Drag blocks into sentence order.
4. Add alt text to figures.
5. Listen to a sample in a screen reader (NVDA is free on Windows, VoiceOver is on Mac).

Merge PDF and Compress PDF typically do not rebuild tags. If the document is regulated, check the output again.

Merge

An accessible cover letter plus an untagged scan appendix is a packet that is only partly accessible. That mix is common. See mixed PDF workflows.

What you can do:

• Tag the digital sections before the merge.
• OCR the scanned sections.
• Run the checker on the merged file.
• Offer HTML or an accessible Word file for a critical form when the PDF cannot be fixed quickly.

Mistakes teams keep making

"We exported PDF, so it is final." A file can look finished on screen and still have no tags.

A scanned form with no fields makes people retype everything. A screen reader cannot help.

Merging without checking page 1 afterward. Tags drop without a warning.

A legal disclaimer saved as an image has to be real text, or read aloud through alt text.

An empty title field hurts orientation in a screen reader. See metadata.

A short check before you publish

• [ ] Text is selectable, or OCR has been applied
• [ ] Headings use styles, or tags exist
• [ ] Reading order has been sanity-checked
• [ ] Images have useful alt text, or are marked decorative
• [ ] Tables have header rows tagged
• [ ] Contrast is readable, not faint gray
• [ ] Merged files were checked again

Privacy

Checkers and repair often run offline on sensitive documents. Prefer a local tool for client material (browser PDF privacy).

When you cannot fix the PDF in time

• Publish the same content as an accessible web page.
• Offer an accessible DOCX next to the PDF.
• Split the packet: an accessible summary PDF, plus appendix scans labeled "image-only appendix."

Related reading

• Word to PDF Conversion Tips
• How to Scan Documents Properly
• Combining Scanned and Digital PDFs
• PDF Metadata: What It Reveals

Limits

A tagged export from Word is a strong start. Complex tables, multi-column callouts, and untagged scan pages still need a pass in a screen reader or a checker such as PAC.

Export with structure from Word, OCR the scans, check again after a merge, and run the list above before a client sees the file. An image-only PDF is not finished.