A PDF you place on your website, for example a manual, a registration form or a price list, is in practice subject to the same accessibility requirements as the web page it appears on. The most common mistake is that such a document was created without a text layer or structure: it is essentially a photograph of text. A screen reader cannot work with it, even though the document looks completely normal to a sighted person.
This error usually does not happen intentionally. A document is scanned and saved as an image, or it is created in Word or InDesign without attention to headings and reading order, and then simply exported to PDF. The website itself may then look fine and be in order in terms of code, while the linked document tells a different story.
What concretely goes wrong
With documents that are not accessible, the problems usually fall into a few fixed patterns. A scanned document without a text layer is the most common: there is literally no text in the file to read or search for. Additionally, a heading structure is often missing, which means that someone using a screen reader cannot navigate through the document the way they can with a properly structured web page. Images, charts and tables in the document often lack a description, and the reading order is not always correct: text that is visually arranged in columns next to each other is then read out of order. A separate issue is the filled-in form as a PDF, where fields have no label and therefore cannot be linked to the question they belong to.
Why this often gets overlooked
Documents are often seen separately from the website. They have been provided by a marketing agency, a printer or another department, and attention is focused on the content and design, not on the structure underneath. Moreover, a scan of a website usually focuses on the pages themselves; the content of a linked document does not automatically receive the same in-depth testing. This does not mean the document should be overlooked, but it does mean it is easily forgotten if no one explicitly looks at it.
How to prevent it
The most practical approach starts with the source file, not the PDF itself. In Word or InDesign you can mark headings as headings (rather than just making them bold or larger), provide images with a description and check the reading order before exporting. Most export functions can carry that structure over to the PDF, provided you use the correct setting. Many office applications nowadays also have a built-in accessibility check that shows before export where something is missing.
For documents that are frequently consulted or are part of an active process, such as an order form or a product manual, it is worth considering offering the content as a regular web page instead of as a download. This saves a separate test and is also more convenient for most visitors. Where a PDF remains the best format, such as for a document that people want to print or keep, the structure in the source file remains the point of action.
What a scan does and does not show
It is good to know that an automated scan can identify some of these problems, for example whether a document is presented as an image without a text layer, but not every requirement lends itself to automated testing. Whether the reading order in a complex document is logical, or whether a description of a chart is factually correct, that requires assessment that goes beyond a technical check. That applies to every scan, including ours, and is therefore noted in every report. how we arrive at our findings explains which checks are automated and where human review remains necessary.
If you want to know where your own website and the documents around it stand, the free quick scan of a few core pages is a logical starting point. For those who want to read more broadly about what is expected of an organization, the overview with all topics by situation provides a short and concrete answer per subject, along with references to the official text. And for individual practical questions that come up regularly, such as about documents, forms or the declaration itself, the collection of answers to frequently asked questions is a place to look first before searching further.
Do you want to know whether this is relevant for your situation and where your site currently stands? The quick scan is free and provides an initial picture within a few minutes, without needing to create an account.