What makes a PDF accessible, and how to fix one that is not

The six things that make a PDF accessible

Accessibility in a PDF is not one setting. It is six separate properties, and a file can have some and not others. Working through them in this order is the fastest route, because each one depends on the ones above it.

Property	What it means	What breaks without it
Real text	The characters are text, not pixels	Nothing can be read at all
Tags	A structure marking headings, lists, tables, paragraphs	Everything sounds like one flat block
Reading order	The order the tags are read in	A two column page is read straight across
Alt text	A description on each meaningful image	Images are skipped silently
Document title and language	Set in the file properties	Announced as the filename; read in the wrong accent
Form fields	Each field labelled, in a sensible tab order	The form cannot be completed

A useful way to think about it: real text decides whether the document exists to a screen reader, tags decide whether it makes sense, and reading order decides whether it makes sense in the right sequence.

Start by finding out whether there is any text at all

This takes five seconds and it decides everything that follows. Open the PDF and try to select a sentence with your cursor.

• The text highlights. There is real text. Go on to tags.
• Nothing highlights, or the whole page highlights as one block. It is a scan, or an image of a page. No amount of tagging will help until it has been through optical character recognition.

If it is a scan, run OCR first. Every serious PDF editor has it, and so do several free tools. Then read the result, because OCR makes mistakes with handwriting, tables, columns and low quality scans, and an accessible document full of wrong words is not much of an improvement.

Tags: the part almost everyone is missing

Tags are an invisible outline inside the file. They say this line is a level one heading, this is a bulleted list, this is a table with these header cells, this is a paragraph. Without them, a screen reader gets a wall of text with no landmarks, so a user cannot skim, cannot jump between sections and cannot tell a heading from a sentence that happens to be in bigger type.

Visual formatting does not create tags. Making a line 18 point and bold makes it look like a heading and tells assistive technology nothing.

The only reliable way to get good tags

Tag the source document, then export. Fixing tags inside a finished PDF is slow, fiddly work; getting them right in Word, Google Docs or InDesign takes almost no time and carries through on export.

1. Use real heading styles in the source, Heading 1 through Heading 3, rather than manually enlarged text.
2. Use the list buttons for lists instead of typing a hyphen and a space.
3. Use a real table with a marked header row, rather than text arranged with tabs.
4. Add alt text to images in the source document.
5. Export with accessibility on. In Word, save as PDF with the document structure tags option enabled. Printing to PDF discards all of it, which is the single most common way an accessible document becomes an inaccessible file.

That last point is worth repeating because it undoes all the other work: print to PDF and you get a picture-like file with no tags. Always export or save as PDF.

Reading order, and the two column trap

Tags say what things are. Reading order says what comes next. They are separate, and a document can have perfect tags in a nonsensical order.

The classic failure is a two column layout. Visually a reader goes down the left column then down the right. If the reading order follows the page geometry line by line, a screen reader reads the first line of the left column, then the first line of the right, and continues alternating. The result is grammatically intact and completely meaningless.

Sidebars, pull quotes, headers, footers and captions cause the same problem in smaller ways: content that visually sits aside gets read in the middle of a sentence. Check reading order on any page that is not a single column, and on every page with a text box.

Alt text: describe the point, not the picture

Alt text is a short description of what an image conveys. The useful question is not what the image shows but why it is there.

• Decorative images such as a divider or a background pattern should be marked as artifacts, so they are skipped. Alt text on decoration is noise.
• Informative images need a sentence that delivers the information. Not "chart", but what the chart says.
• An image of text needs the text. Scanned signatures, letterheads and screenshots of tables are common offenders.
• Logos need the organisation name, not the word logo.

A chart is worth a special note. Alt text cannot carry a whole dataset, so give the conclusion in the alt text and put the numbers in a real table nearby, which helps everyone and makes the figures usable.

Title, language and the small properties nobody sets

Two settings in the document properties take a few seconds and are missing from most files.

• Document title. Without it, assistive technology announces the filename. "final-v3-approved-2.pdf" is not a document title. Set a real one and set the file to display it.
• Language. A screen reader uses it to choose pronunciation. An English document with no language set may be read in whatever voice the user has configured, and a French document read with English pronunciation is close to unusable.

If a document mixes languages, mark the passages individually so each is pronounced correctly.

Forms are the hardest part, and usually the wrong format

An accessible PDF form needs every field to have a label that assistive technology reads, a tab order matching the visual order, grouped radio buttons and checkboxes, and error and instruction text available as text rather than as a red outline.

That is achievable and it is a lot of work. Before doing it, ask whether the thing needs to be a PDF at all. A web form is easier to make accessible, works on a phone, can validate as the person types, can be saved halfway through and does not require anyone to find a PDF editor.

Situation	Better format
You collect information from people	A web form, almost always
A signature is needed	A web form plus an e-signature step
A specific printed layout is required	PDF, tagged properly
A government or regulator mandates the form	Their PDF, with an accessible alternative route offered
A document is for reading, not completing	A web page, with a PDF as a secondary copy

The last row is the one that saves the most effort over time. A web page is accessible by default if it is built reasonably, stays correct when you edit it, and is easier to find. A PDF needs the work done again every time the document changes.

When you are on the receiving end of a PDF form that is not accessible and you have no choice about it, you can at least avoid printing it: open it with the fill a PDF tool, type into the fields, and add a signature with sign PDF. That does not make the file accessible, and it does mean you are not forced to find a printer and a scanner.

How to check what you have

1. Select some text. If you cannot, it is a scan and needs OCR first.
2. Run the built-in accessibility check in your PDF editor. It finds missing tags, absent alt text, no title and no language quickly.
3. Read the tag tree and confirm headings are headings and tables have header cells.
4. Check reading order on every page that is not a single column.
5. Tab through a form and confirm the focus moves in the order a person would read it.
6. Listen to a page with a screen reader. This is the only step that tells you whether the document actually works.

Step two finds the mechanical problems and will report a clean file that is still unusable, because no checker can tell whether your alt text is accurate or your reading order makes sense. Step six is the one that finds the real failures. Our guide to checking PDF accessibility goes through each of these in order with what to look for.

If you have hundreds of PDFs

Remediating an archive one file at a time rarely finishes. Triage instead.

• Forms people must complete: rebuild as web forms. Highest value by a wide margin.
• Current, frequently downloaded documents: remediate properly, in order of how often they are opened.
• Key information published only as a PDF: publish as a web page and keep the PDF as a secondary copy.
• Old archives nobody opens: leave them, and offer an accessible version on request.
• Everything you produce from now on: fix the templates. This is what stops the problem growing.

The last item decides whether this is a project with an end. If your templates use real heading styles, marked table headers and alt text, accessible documents fall out of ordinary work and nobody has to remember anything.

Next step

Pick your most downloaded PDF and try to select its text. That one test tells you which half of this guide applies to you. If it is a form people fill in, the best fix is not to tag it better but to stop sending a PDF.

If you are stuck completing somebody else's inaccessible form today, you can do it in your browser with the free PDF filler and sign it with sign PDF, without printing anything. The file stays on your own device.