Check reading order in a PDF

A two-column article often needs the whole left column before the right. A comparison built around paired rows may need a different sequence. Do not impose the same mechanical Z-pattern on every page.

Tag order is central, but inspect text within tags and practical reading too. The tab order of interactive objects is another separate check.

Sources: [1]

Take a page with two columns and a fact box. First establish which main-text paragraph continues which. Then place the box where it adds information without breaking a sentence or argument. The intended sequence might be the main text, the box and the next section. On another page, the box may need to come first to explain a concept.

There is no universal rule to read everything geometrically from left to right and top to bottom. Tables are an obvious exception: readers need row and column relationships, not merely every word in spatial order. Start with meaning and purpose.

Write a simple named sequence, such as “heading, introduction, left column, right column, background box”. Use it when inspecting the structure. If editors cannot agree on the sequence, the layout or wording itself may need clarification.

• Write the intended reading sequence for each page type.
• Follow tags and locate their visible page content.
• Read continuously with assistive technology and note interruptions.
• Inspect page breaks, notes, sidebars and figures.
• Test the keyboard sequence for links and fields.

The tag tree describes logical structure, but one tag can contain several text fragments. Inspect the sequence within longer paragraphs and objects too. A correctly positioned P tag does not help if its text jumps between columns.

Keyboard tab order concerns interactive objects such as links and form fields. It must support a coherent task but is not the order of every body paragraph. If someone can read an instruction but Tab then moves to the wrong field, investigate that interaction separately.

Text extraction, reflow and assistive-technology presentation may yield different observations. Use them diagnostically and state which method produced each result. A copying test can suggest a problem subsequently confirmed by structural inspection and practical reading.

After repair, revisit transitions: before and after a box, across a page break and between tables and prose. Check that repeated headers do not interrupt content and meaningful notes remain discoverable. These joins often make an apparently correct page difficult to read.

Copying all text into a text document can reveal problems, but is not a complete screen reader test. Applications may use different information for text extraction. Combine methods instead of allowing one view to settle the result.