A two-column handout can look immaculate in a PDF and be hard to read when you listen to it. The eye follows the left column to its end, then moves to the right. A reading tool may instead encounter a heading, a sentence from the other column, a decorative note and the rest of the original sentence. The page has preserved enough appearance to satisfy a quick visual inspection. The person using it needs a sequence, relationships and a way to navigate the contents.
This is why it is worth testing the exported file, even if the web page or source document already works well. A PDF can carry a logical structure in addition to its visible marks. Tags can identify headings, paragraphs, figures and other kinds of content. Reading order determines how related pieces are encountered. Interactive elements also have an order of focus. The W3C’s PDF3 technique describes ways to verify reading and tab order in a tagged PDF, including listening to the content with assistive technology and checking the order of focusable elements. W3C, PDF3 on tab and reading order
Start with the information the reader needs to recover
Imagine a three-page handout announcing a workshop. A heading says who it is for. A two-column section gives a schedule and instructions for the location. A small box on the first page describes an access arrangement. A link near the end opens the booking form. The layout works for somebody scanning the page visually because they can see which details are grouped. For someone reading in sequence, the links between those pieces need to survive in another form.
Before exporting, write down the expected reading order for one difficult page. Keep the description short. A title introduces the event, an overview explains what it is, and the schedule and access information follow in a deliberate sequence. A booking link belongs where the reader has enough information to decide whether to use it. When you listen to the resulting PDF, the written expectation tells you what to check. Without it, it is easy to recognise that something feels confusing without knowing which relationship the file has lost.
Then inspect the source. A line that looks large and bold may be a heading in a reader’s mind and plain text in the file. A table made of individually positioned blocks may look orderly while giving the export process little evidence about its rows or header cells. A chart beside a paragraph may depend on a caption whose position is obvious visually and unclear in a text sequence. Apply the meaningful structure in the authoring environment before relying on the export settings. Microsoft describes the use of accessibility features in source files and settings for preserving document structure when exporting a PDF. Microsoft Support, creating accessible PDFs
Keep one important distinction visible. Text that can be highlighted or copied indicates that the file has usable text in some form. It does not establish that headings can be reached through heading navigation or that the contents are in the right order. Conversely, an individual reader’s successful trip through a simple page does not verify every table, link and figure in a larger document. Treat each test as an answer to a specific question.
The route to PDF affects what arrives
A PDF made through an application’s export command and one made through an operating system’s printer can follow different paths. Browsers may have their own save destination, and a workflow automated through a browser can expose a separate option for tags. The Chromium project announced tagged PDF output for its Save as PDF destination beginning with Chrome 85. The DevTools Protocol lists an option for generating tagged output in its print-to-PDF command. These facts establish that some printing paths can carry structure. The particular software, settings and document still determine what is present and useful in the file you deliver. Chromium, tagged PDFs in Chrome · Chrome DevTools Protocol, Page.printToPDF
If you maintain a document pipeline, record how the PDF was produced. “Saved from the browser” is too broad for a colleague who has to reproduce the result. Identify the browser and route, the source revision and any print styling that affects what appears. If an authoring application is used, record the export method and the structure options that apply. For an automated process, keep the relevant configuration with the job. A reproduction attempt should be able to reach the same format without depending on the undocumented settings of one computer.
Use a representative test rather than an empty template. A one-page text sample may pass while a full document fails on the first two-column section or multipage table. Include one heading, at least one link, a figure with a meaningful description and a layout where order is easy to lose. If a component is supposed to be omitted in print, check that the omission is intentional and that no instruction depends on it. A document that tells a reader to press a button that disappeared during export has a content problem even if the remaining elements are correctly tagged.
When the source is a web page, print styles can influence what survives and where it appears. They might hide menus, alter columns or introduce page breaks. Those can be sensible changes, but test the result as an independent document. A caption that follows a figure on screen can land on another page. A note intended for one section can become separated from it. Check whether the document still provides enough context for a reader who only has the exported version.
Listen to an actual difficult passage
Open the exported PDF using the software available for your test. Listen to the first page and the passage whose order is most likely to be lost. Try moving by headings if the reading software supports it. Check whether headings identify sections, whether the transitions make sense and whether the figure description appears where it is needed. The W3C technique gives examples of manual verification using a screen reader or another reading tool and describes inspection through accessibility interfaces. W3C, PDF3 verification procedures
Then examine keyboard focus if the PDF contains links or form controls. Move forward and backward between them. Ask whether the order matches the task a reader needs to complete. In the workshop example, a link to travel directions might make little sense before any location has been identified. In a form, a focus jump from a name field to the final submit action can make the rest of the fields difficult to discover. The fact that you can eventually reach every element is one useful observation. The order in which you reach them is another.
Look for repetition as well as omissions. A decorative icon announced several times can interrupt a sentence or make similar sections hard to distinguish. A caption repeated in a figure description and in nearby body text may make a clear page sound laborious. Review these cases in context. Sometimes the description needs to explain the information conveyed by the image, while the visible caption identifies its subject and date. Sometimes the same words are intentional because the two mechanisms reach different readers. The goal is a coherent passage in the delivered file.
A table needs a more targeted check. Pick a value in the middle and ask which column and row identify it. A reading tool that speaks a series of numbers without the corresponding labels may leave the reader unable to tell whether a number is a time, a price or a capacity. Verify that the relevant headers are associated with the data in the structure. The W3C’s collection of PDF techniques includes approaches for tables and other semantic features, separate from the technique for reading order. W3C, PDF accessibility techniques
Put corrections where they will survive the next export
Check the language and title alongside the order tests. A descriptive document title gives someone a useful name in a window or file list, while language information can affect the way a reading tool pronounces the text. In a bilingual handout, inspect how each language is represented and listen to the transition between them. The fact that the letters are rendered correctly on screen tells you little about that transition. You can also inspect whether a meaningful link describes its destination when reached on its own. Several links all called “click here” are hard to distinguish in a list even if their positions in the running text make sense. These checks answer different questions from the presence of tags or the page’s reading sequence.
If the schedule jumps between columns, try to establish why before repairing the final file. The source may be organised as visually positioned fragments. The export may have produced a structure different from what you expected. A later editing operation may have altered it. Checking the source order and the exported tags can distinguish these cases. If the authoring file can express the intended structure, fixing it there makes the next revision more likely to retain the repair.
A final-file repair can still be appropriate when the source is unavailable or when a particular export requires adjustment. Adobe describes Acrobat Pro’s Reading Order tool as a way to inspect and change aspects of order and tagging. Keep a record of any changes made only to the PDF. When somebody updates the source or replaces a page, they need to know which repairs may need to be repeated. Verify the repaired file using the same difficult passage and keyboard steps as before. A change made in an editor is a proposal until the delivered file has been checked. Adobe, Reading Order tool for PDFs
For a document with recurring editions, keep the test narrow enough to repeat. After every export, check that the page most likely to break still reads in order and that any link needed for the main task is reachable. When a new layout, table or form component is introduced, add a test for that element. A record of the exact source revision, export method and verified file helps somebody identify which version was actually inspected. It also makes a report of a problem easier to reproduce.
If an equally useful HTML version is available, keep it easy to find alongside the PDF and make sure it receives the same content corrections. People may have different reasons for wanting a fixed document or a web page. The work on the source benefits both when the organisation of the information is clear. It remains necessary to test each version on its own because page layout, navigation and export can change how readers encounter the content.
In the workshop example, the result is good when a reader can find the schedule, connect the access information to the right part of the visit and reach the booking link in a sensible order. The PDF may look the same before and after the correction. The difference appears when someone tries to use it. That is the comparison to keep for the next document you export.