18 PDF Checks to Run Before You Publish

Confirm the file has real text

Try selecting a word with your cursor, or search for one. If nothing selects, what you have is a photograph of a page and there is no text in it to read out. That file needs optical character recognition run over it before any other check on this list means anything.

Confirm the file has a tag tree

Run the free checker, or open the tags panel in Acrobat Pro and look for content under it. An empty panel means the document was exported without structure, which is the single most common state we find. Retagging from scratch is slow. Fixing the source file and exporting again is usually cheaper.

Set a real document title

Find the title field in the document properties and write what the document is, so "Spring 2026 price list" rather than a filename with underscores in it. This is the same rule that governs a web page title, and it is announced first when the document opens.

WCAG 2.4.2

Show the title instead of the filename

A separate setting in the initial view options decides whether readers see the title you just wrote or the filename. Setting the title without turning this on changes nothing anybody sees, which is why these two checks are always done together and why doing only the first is such a common near miss.

WCAG 2.4.2

Set the document language

The language setting in the advanced properties decides which voice and pronunciation a screen reader uses for the whole file. It takes one selection. Getting it wrong does not make the reading slightly worse. It makes it unintelligible.

WCAG 3.1.1

Tag headings at the right levels

Open the tags panel and read the heading levels down the document. They should describe the shape of the document rather than the size of the type, so a subsection under a section is a level below it. Gaps in the sequence are worth closing for the reader, and they are advice rather than a rule, so fix them as a defect rather than logging them as a breach.

Check the reading order end to end

Select all, copy, and paste into a plain text editor. What you get back is a fast read on the order the content sits in, and it is a diagnostic rather than a verdict, because copy order and tag order are not the same thing. Scrambled text means go and look at the tag tree. A two-column layout pasting as one interleaved column is the classic export failure.

Tag lists as lists

A real list uses list tags around list items, not a run of paragraphs each starting with a hyphen. The difference matters because a screen reader announces a real list with its length, so somebody knows there are seven items coming and can skip past them.

Place captions and sidebars in the flow

A caption belongs directly after the thing it describes, and a pull quote or sidebar belongs where a reader would meet it. Exporters commonly float both to the end of the page, which puts a caption several paragraphs after its figure and leaves it stranded.

Give meaningful images alt text

Find the figure tag, open its properties and write the alternate text. Describe what the image tells a reader at this point in this document, which is not always what is visible in it. The same photograph needs different words in a case study and in a staff directory.

Mark decorative images as artifacts

Dividers, borders, watermarks and background textures should be marked as artifacts, which takes them out of the structure tree and out of the reading order. It is the PDF equivalent of an empty alt attribute rather than another kind of tag. Skipping it leaves a reader hearing "graphic" every few lines with no way to tell which ones matter.

Describe charts fully

Naming a chart tells somebody what they are missing without giving it to them, and the rule asks for an alternative serving the same purpose. So give the figure a short alt naming what it is, then put the actual figures into the document as real text near it, where every reader gets them.

Tag tables as real tables

A table needs table, row and cell tags around it, not a grid of separately positioned text blocks that happens to line up. Check the tags panel rather than the page, because a table drawn with tabs and spaces looks identical on screen and carries no relationships at all.

Mark header cells as headers

Set the cell type to header on your top row, and on the first column too where that column labels the rows. This is what lets a screen reader read "Revenue, Q3, 14.2" instead of a bare number, and it is the single change that makes a table usable by ear.

Set scope on headers where the table needs it

Scope tells a cell whether its header governs a row or a column, which matters as soon as a table has both. Worth knowing where this sits, because W3C's own table technique asks only that header cells are tagged, so scope is a document-standard requirement rather than a WCAG one. It still earns its place on anything larger than a simple grid.

Name every form field

Each field's tooltip becomes its accessible name, so the visible label text has to appear inside it. The rule asks the name to contain what the eye can see rather than to match it exactly, so extra words in the name are fine and missing ones are not. That matters more than it sounds, because somebody using voice control says what they can see, and a field labeled Surname whose name is "lastname2" cannot be reached that way.

Mark required fields as required

Tick the required option in the field's properties, so a screen reader announces the field as required at the moment somebody arrives in it. An asterisk is fine alongside that, as long as a legend on the form explains what it means. What the flag does not do on its own is tell somebody what went wrong, so test the error handling separately.

Write link text that says where it goes

Not a bare address, which gets read out character by character in places, and not "click here", which tells a reader nothing once it is out of its sentence. Say what is at the other end, and add the file type and size where the link leads to another document.