What makes a PDF accessible? Tags, reading order and alt text

The short version

A PDF looks the same to everyone on screen. It does not sound the same. People who use screen readers, magnifiers, braille displays or read-aloud tools depend on information hidden inside the file: what each block of text is, which order to read it in, and what each picture means.

Adobe describes the core of it plainly: document structure tags "define the reading order and identify headings, paragraphs, sections, tables." If that layer is missing or wrong, the file may look fine and still be unusable for some readers.

Here are the parts that make the difference, in the order most checkers look at them.

1. Real text, not a picture of text

A scanned page is an image. A screen reader has nothing to read. Adobe notes that scanned images of text are inaccessible until they go through OCR (optical character recognition), the step that turns letter shapes into actual characters.

How to tell: try to select a single word. If the whole page highlights as one block, or nothing highlights, the page is an image.

W3C lists this as its own technique for PDFs: performing OCR on a scanned PDF "to provide actual text" (technique PDF7).

2. Tags: the hidden structure

Tags are labels stored inside the PDF that say what each piece of content is. A tagged file knows that one line is a heading, the next is a paragraph, and the block below is a list with five items.

The common ones:

• Headings (H1 to H6): let screen reader users jump from section to section, much like a table of contents.

• Paragraphs: ordinary body text.

• Lists: tell the listener how many items there are and where the list ends.

• Tables: mark which cells are headers, so each number is read with its row and column label.

• Figures: images, charts and diagrams, each with alt text.

• Links: with text that says where they go.

W3C's guidance for PDFs (part of the techniques for WCAG, the Web Content Accessibility Guidelines) covers each of these separately: heading tags (PDF9), list tags (PDF21), table markup (PDF6) and link structure (PDF11).

3. A logical reading order

The reading order is the sequence a screen reader follows. In a two-column newsletter with a sidebar and a pull quote, the correct order is not always top to bottom, left to right. A poorly tagged file can jump from the first column straight into the sidebar, then back.

W3C's technique PDF3 recommends testing this by listening to the document with a screen reader or a read-aloud tool and confirming the order makes sense.

4. Alt text for meaningful images

Alt text (alternative text) is a short description stored with an image. Adobe puts it simply: images and form fields "cannot be understood by the user of a screen reader unless they have associated alternative text."

Good alt text says what the image is for, not just what it shows. For a sample chart in a fictional Northgate County budget report:

• Weak: "Bar chart."

• Better: "Bar chart: road maintenance is the largest department budget, about twice the size of parks."

If the same numbers sit in a table right next to the chart, the alt text can say so and point there.

5. Decorative items marked as artifacts

Not everything on the page should be read aloud. Page numbers, running headers, background shapes and divider lines are usually decorative. In PDF terms these are marked as artifacts, which tells assistive technology to skip them. W3C covers this in technique PDF4, "Hiding decorative images with the Artifact tag." Without it, a listener can hear "Page 4 of 38, Northgate County Annual Report" at the top of every page.

6. Document title and language

Two small settings that checkers flag constantly:

• Title: the file should show a real title (for example, "Northgate County Budget Summary 2026") rather than a file name like "final_v3_scan.pdf." Adobe notes that a title helps users identify the document. W3C covers it in PDF18.

• Language: setting the document language helps a screen reader pronounce words correctly (W3C technique PDF16). If one passage is in another language, it can be marked separately (PDF19).

7. Forms, links and bookmarks

If the PDF is a form, each field needs a label a screen reader can announce, and required fields should be marked as required (W3C techniques PDF10 and PDF5). Longer documents benefit from bookmarks (PDF2), which give everyone a clickable outline.

Link text should make sense on its own. "Read the 2026 budget summary" is clearer than "click here."

8. Color and contrast

Text needs enough contrast with its background, and color should not be the only way to show meaning (for example, red for overdue with no label). These checks apply to PDFs just as they do to web pages.

Which standards describe this

• WCAG is the W3C standard most laws and policies point to. W3C says the current version is WCAG 2.2, published in October 2023, and that WCAG "can also be applied to non-web" documents.

• PDF/UA (ISO 14289-1) is the PDF-specific standard for universal accessibility. Adobe lists it alongside WCAG as guidance for accessible PDFs.

Which version and level an organization must meet depends on the law or policy that applies to it, which varies by country and sector.

A quick self-check before you publish

1. Can you select the text?

2. Does the file have tags at all? (Most PDF readers show this in document properties or a Tags panel.)

3. Do the headings form a sensible outline, with no skipped levels?

4. Does each meaningful image have alt text that explains its purpose?

5. Are tables marked with header cells?

6. Are page numbers and running headers marked as artifacts?

7. Are the title and language set?

8. Does a read-aloud tool read the page in the order a person would?

Automatic checkers catch some of these. They cannot judge whether alt text is accurate or whether the reading order makes sense, so a person still has to look.

Get a head start on your own PDF

EasyToDecode is launching soon with a tool for exactly this job. You upload a PDF and get back a tagged version with headings, reading order and table structure, draft alt text for you to review, and a report of what still needs a person. It does not certify compliance, and it tells you which items to confirm. See how it works or join the waitlist to hear when it opens.