# Review how the exported PDF can actually be read

Complete original constructed teaching cases. No real people, studies, source datasets, customers, accounts or external publications.

Review the saved PDF’s structure, reading and tab order, navigation and document context, then record actual reader-tool checks separately. Start with a well-structured source and inspect the export again after repair. A clear preview, selectable text, bookmark list or “tagged” flag alone does not establish an accessible reading path or WCAG/PDF-UA conformance.

## Identify the file and the promised reading task

Keep the actual public export, its source version, language, intended reader task and any permitted companion format together. Open the saved file in the intended viewer. An editor canvas can show a different version, and a native platform’s preview can differ from the file readers download. This guide concerns structured reading review; privacy removal and native upload are separate tasks linked below.

Use semantic structure in the source where the authoring tool supports it, then inspect what survived conversion. Do not assume a paid tool is necessary for every workflow or that an export option certifies the output. W3C’s PDF techniques illustrate ways to meet particular requirements; the techniques themselves are not mandatory rules or a complete conformance checklist.

## Compare logical meaning, structure order and focus order

Write the intended reading sequence in plain text before inspecting tags: title, context, first input, calculation, qualification and next step. W3C PDF3 explains that reading order depends primarily on tag order, with content-tree order within individual tags. A visually obvious two-column layout can therefore become an interleaved explanation when the structure is wrong. A “tagged” flag says too little about that sequence.

Focusable links or form fields need a usable tab order as well. Read the whole relationship: a reader should encounter the necessary instruction before the action and be able to leave the control. A text extractor, tag-tree inspection, keyboard path and actual screen-reader reading each reveal different properties. Keep a specific failure and its saved file identity instead of collapsing them into one pass.

## Check navigation and context in the saved document

Inspect actual heading tags and hierarchy, not only large bold text. PDF9 describes heading structure that assistive technology can expose for navigation. Bookmarks provide a different outline route: PDF2 treats them as an advisory technique for locating content in longer or subdivided documents. Check both the displayed entries and their actual destinations. A bookmark does not repair a paragraph incorrectly marked as a heading.

Check document language and a meaningful title independently. PDF16 describes the document’s default /Lang entry. PDF18 describes /Title in the document information dictionary and a separate DisplayDocTitle viewer preference. A filename is not the title metadata, a title does not set the language, and a language setting does not translate the text. Reader software can expose these settings differently, so record the actual saved-file observation.

Compare visible page labels with viewer navigation. In our constructed four-page example, the labels are i, ii, 1, 2: physical page three is labeled 1. PDF17 describes /PageLabels for that distinction. A viewer showing only physical “3” when the instruction says page “1” can misdirect the reader. Check the relevant section boundary and link destination, not merely the total number of pages.

| Physical page | Visible label | Expected navigation target |
| --- | --- | --- |
| 1 | i | 1 |
| 2 | ii | 2 |
| 3 | 1 | 3 |
| 4 | 2 | 4 |

## Resolve three structured-reading failures

P1 — Constructed column order: the intended sequence is A context, B inputs, C result, D limitation. The visible layout looks correct, but the inspected structure yields A, C, B, D. Decision: repair the source/export structure or the saved tags with suitable existing software, then repeat the actual reading-order check. Outcome: a declared failure and repair plan; selectable text or a tagged flag does not make the wrong order a pass.

P2 — Constructed page navigation: four physical pages have visible labels i, ii, 1, 2, while the viewer and a saved reference use physical page three as “3.” Decision: align the document’s page-label information and reference wording, then check that the requested label 1 reaches physical page three. Outcome: a four-page mapping exercise, not an executed PDF edit or proof of every viewer’s behavior.

P3 — Constructed incomplete review: language and title are present and bookmarks open, but heading navigation, meaningful sequence and the intended assistive-technology path have not been inspected. Decision: record those checks as open and supply the essential explanation in an appropriate readable companion. Outcome: partial technical observations with a pending reading review, not a full accessibility badge or proof that the companion automatically resolves every barrier.

## Use an actual reading check and retain its limits

For an authorized review, record the file, viewer/assistive technology and version, task, actual route and observed result. Try locating a heading, following the worked explanation, reaching a relevant link and returning to context. Follow W3C’s relevant procedure with suitable tools and actual observations. Do not fill a screen-reader result with a text-extraction log or a simulated reader.

Repeat affected checks after changes, using the saved export and any obtainable destination copy. State remaining gaps such as an untested viewer, image meaning, forms, contrast or reflow. A structured-reading review is valuable but narrower than a full PDF, accessibility or legal assessment. When an essential path remains unusable, hold the affected distribution or provide a genuinely usable route and describe the remaining limitation.

## Save the exact checks without inventing certification

The private record includes source/export identity, intended sequence, inspected tags and focus order, heading/bookmark destinations, language/title/page-label observations, actual reading-tool route, repair and retest, companion and unresolved checks. The examples provide complete source/observed-order/decision traces. The local checker tests the fictional sequence and page-label mappings; it does not open or certify a PDF.

The six W3C technique bodies were read on 1 November 2026. PDF2’s web-tool observation timed out but its complete direct body was read successfully. These are informative techniques with criterion-specific scope. No Acrobat, screen reader, real exported PDF or native upload was operated for the fictional P-cases. They are not a new research study, WCAG/PDF-UA conformance statement or independent user test.

1. Identify the saved public file.
2. Compare intended meaning with structure and focus order.
3. Check headings and bookmark destinations.
4. Inspect language, title and page labels.
5. Observe an actual reading-tool path and retest after repair.

## Primary sources

- [W3C PDF3: Logical reading and tab order](https://www.w3.org/WAI/WCAG22/Techniques/pdf/PDF3)
- [W3C PDF9: Tagged headings](https://www.w3.org/WAI/WCAG22/Techniques/pdf/PDF9)
- [W3C PDF2: Bookmarks and destinations](https://www.w3.org/WAI/WCAG22/Techniques/pdf/PDF2)
- [W3C PDF16: Default document language](https://www.w3.org/WAI/WCAG22/Techniques/pdf/PDF16)
- [W3C PDF17: Page labels and navigation](https://www.w3.org/WAI/WCAG22/Techniques/pdf/PDF17)
- [W3C PDF18: Document title and viewer preference](https://www.w3.org/WAI/WCAG22/Techniques/pdf/PDF18)