10 min read
How to check whether a PDF is accessible
Checking a PDF for accessibility is not one test, and the automated checker that comes with your PDF editor is the least informative of the six. It finds missing tags and absent alt text, which is useful, and it will happily pass a document that is unusable because no software can judge whether your alt text is accurate or your reading order makes sense. This guide gives the six checks in the order that finds problems fastest, with what a pass and a failure look like at each step. It is general guidance rather than legal advice.
The six checks, in order
Run them in this order. Each one is cheap and each tells you whether the next is worth doing, so a file that fails step one needs OCR before anything else is meaningful.
| Step | Check | Time | Finds |
|---|---|---|---|
| 1 | Select the text | 5 seconds | Whether it is a scan |
| 2 | Automated accessibility check | 1 minute | Missing tags, alt text, title, language |
| 3 | Read the tag tree | 5 minutes | Fake headings, tables without headers |
| 4 | Check reading order | 5 minutes | Columns read across, stray text boxes |
| 5 | Tab through the form | 2 minutes | Unlabelled fields, wrong tab order |
| 6 | Listen with a screen reader | 10 minutes | Everything the others miss |
Step 1: try to select the text
Open the file and drag your cursor across a sentence.
- Individual words highlight. There is real text. Continue.
- Nothing highlights. It is an image. Stop and run optical character recognition, then start again.
- The whole page highlights as one rectangle. Also an image. Same answer.
A second quick test: use your reader's find function to search for a word you can see on the page. If it is not found, the word is not there as text.
When you do run OCR, proofread it. Optical character recognition is confident and frequently wrong about handwriting, poor scans, tables and anything in columns.
Step 2: run the automated check
Every serious PDF editor has an accessibility checker, and several free tools will do it. Run it and fix what it reports. It is good at exactly the mechanical absences: no tags, no document title, no language, images with no alt text, tables with no header cells, form fields with no label.
Two things to keep in mind. A checker reports what is missing, not what is wrong: it cannot tell that your alt text says "image1" or that your heading structure skips from level one to level four. And a clean report is not a pass, which is why there are four more steps.
Step 3: read the tag tree
Open the tags panel. You are looking at the document as assistive technology sees it, and three questions answer most of it.
- Are the headings real headings? Each visual heading should appear as a heading tag, at a level that reflects the outline. Text that looks like a heading but appears as a paragraph is the most common fault in the whole exercise.
- Do the levels descend sensibly? Level one, then two, then three. Jumping from one to four leaves gaps that screen reader users navigate by.
- Do tables have header cells? A data table needs its header row marked as headers, otherwise every cell is read without the context of which column it is in.
Also check that decorative images are marked as artifacts rather than carrying alt text, and that no real content has been marked as an artifact, which hides it completely.
Step 4: check the reading order
Reading order is separate from tags, and this is where visually tidy documents fail. Open the order or content panel and step through the page.
- Multi-column pages: confirm the order goes down the first column and then down the second, not across the page line by line.
- Text boxes, sidebars and pull quotes: confirm they are read somewhere sensible rather than interrupting a sentence.
- Headers, footers and page numbers: these should usually be artifacts, not read on every page.
- Captions: should sit next to the image they describe.
If a document is a single column of body text with no boxes, this step is usually fine. Anything designed is worth checking page by page.
Step 5: tab through any form
If the file has fields, put the mouse down and press Tab repeatedly from the top.
- Does focus start at the first field a person would fill in?
- Does it move in the order the form reads, rather than the order the fields were created?
- Can you tell where focus is at every point?
- Are radio buttons and checkboxes grouped so their question is announced with them?
- Can you reach and activate the submit or signature field?
Tab order is set by the field order in the file, not by position on the page, so a form edited over time usually has an order nobody intended. This test finds it in two minutes.
Step 6: listen to it
This is the step that tells you whether the document works, and the one people skip. Use a screen reader and listen to one or two pages. VoiceOver is built into macOS and iOS, Narrator into Windows, and NVDA is free.
You are not auditing. You are listening for four things.
- Does the document announce a real title, or a filename?
- Does it make sense in the order it is read?
- Are headings announced as headings, so you could skim by jumping between them?
- Do images convey anything, or are they silently absent?
Ten minutes of this is worth more than any report. It is also the only test that reflects what someone relying on the document actually experiences.
What to do with the result
Most files fail on tags, reading order or both, and fixing them inside a finished PDF is slow. If you still have the source document, it is almost always faster to fix the source and export again than to repair the PDF. Our guide to accessible PDFs covers what to change in the source and how to export without losing it.
If the file is a form people have to complete, the most useful outcome of this test is often the decision to stop using a PDF form. A web form is easier to make accessible and easier to keep that way.
And if you are simply trying to complete an inaccessible form that somebody else sent you, you do not need to fix it: open it with the free PDF filler, type into the fields, and add your signature with sign PDF. Nothing is uploaded and nothing is printed.
Frequently asked questions
How do I check if a PDF is accessible?
Run six checks in order: try to select the text, run your editor's automated accessibility check, read the tag tree, check reading order on multi-column pages, tab through any form fields, and listen to a page with a screen reader. The last one finds what the others cannot.
Is a clean automated accessibility report enough?
No. A checker finds what is missing, such as absent tags or alt text, but it cannot tell whether your alt text is accurate, whether a visual heading is tagged as a heading, or whether the reading order makes sense. It will pass documents that are unusable.
What is the tag tree and why does it matter?
It is the invisible structure inside the PDF marking headings, lists, tables and paragraphs. It is how assistive technology understands the document, and visual formatting does not create it: making text big and bold does not make it a heading.
Why is reading order separate from tags?
Tags say what each element is; reading order says what comes next. A two-column page can have perfect tags and still be read straight across the page, alternating between columns, which produces grammatical nonsense.
What free tools can I use to test a PDF?
Most PDF editors include an accessibility checker. For the listening test, VoiceOver is built into macOS and iOS, Narrator into Windows, and NVDA is a free screen reader for Windows.
Can I fix accessibility in the PDF, or should I fix the source?
Fix the source and export again whenever you still have it. Repairing tags and reading order inside a finished PDF is slow and fiddly, while getting headings, lists, tables and alt text right in the source document takes very little time.