Does a clean automated accessibility report mean a document is accessible?

No. Automated checks can identify many structural defects, but they cannot reliably judge meaning, accurate alternatives, logical reading order, or every interaction. Combine them with manual inspection and relevant assistive technology tests.

Should remediation happen in the source file or the final PDF?

Repair the source when it is available and suitable, particularly for documents that will be updated or regenerated. Remediate the final PDF when necessary, and verify the actual deliverable regardless of where repairs were made.

Does OCR make a scanned document accessible?

OCR creates recognized text, which must be checked for accuracy. The document still needs appropriate structure, reading order, language information, image equivalents, and testing.

Why does ADS include training and policy alongside technical document work?

The ADS domains include organizational workflows, training, support, and policy advice because reliable accessibility depends on how documents are created, approved, regenerated, and maintained, as well as how individual files are repaired.

Accessible Document Specialist (ADS): 60 Concepts

This guide supports intermediate-level preparation for the IAAP Accessible Document Specialist (ADS) Examination. Its concepts follow the five knowledge domains, moving from accessible authoring to remediation, testing, organizational planning, and policy advice. Use the explanations, worked examples, and specific mistakes to connect document structure with the experience of people using assistive technologies across word processing, spreadsheet, presentation, PDF, and EPUB formats.

Creating Electronic Documents

1. Plain language and Easy Read

Plain language helps intended readers find, understand, and use information through clear wording, organization, and design. Easy Read is a distinct approach that commonly combines simplified text with supporting images and careful layout. Choose the approach for the audience and purpose rather than treating shorter words as a complete accessibility solution.

Worked example: A notice says, “Prior authorization is requisite.” Rewriting it as “Get approval before booking” preserves the instruction while making the required action clear.

Mistake to avoid: Removing an essential condition while simplifying a sentence changes its meaning.

2. The four accessibility principles

Perceivable, operable, understandable, and robust describe complementary accessibility needs. Information must be available through usable forms, interaction must be possible, meaning and operation must be understandable, and structure must support compatible technologies. A document can satisfy one principle while failing another, so evaluate them together.

Worked example: A form has readable text but its submit control cannot receive keyboard focus. Its visible information is perceivable, but the submission task has an operability barrier.

Mistake to avoid: Treating readable appearance as evidence that every document interaction is accessible.

3. Semantic heading hierarchy

Heading styles communicate section relationships and enable navigation; font size alone does neither. Assign levels according to the document outline, with subordinate sections beneath their parent sections. Headings should describe their content and make sense when encountered in a navigation list without surrounding paragraphs.

Worked example: Under the main heading “Volunteer Handbook,” use a second-level heading “Expenses” and a third-level heading “Travel claims.” A separate second-level heading “Safety” begins the next main section.

Mistake to avoid: Choosing heading levels because their default visual formatting looks attractive.

4. Lists that communicate relationships

Use the authoring tool's list functionality to encode related items. Ordered lists communicate meaningful sequence; unordered lists communicate a collection without required order. Nested lists express subordinate relationships. Typed numbers, bullets, or repeated tabs may resemble a list while leaving assistive technology with unrelated paragraphs.

Worked example: An application process uses a numbered list for “Complete the form,” “Attach evidence,” and “Submit.” A separate bulleted list names acceptable evidence because those items have no required sequence.

Mistake to avoid: Using manually typed numbering that loses list semantics during export.

5. Data tables and header relationships

A data table represents relationships between rows and columns. Identify header cells and keep the structure as simple as the information allows. Readers need enough context to understand each value without relying on its visual position. Use layout tools rather than tables when the purpose is merely arranging unrelated content.

Worked example: A table has columns “Department” and “Requests.” Marking the first row as headers lets the value 14 be understood as the number of requests for the department in that row.

Mistake to avoid: Bold formatting does not, by itself, identify a cell as a table header.

6. Spreadsheet orientation and readability

A spreadsheet needs clear worksheet names, identifiable data regions, and meaningful row and column labels. Wrapping text and adjusting dimensions can expose information otherwise hidden visually. Avoid merged cells within data regions when they disrupt navigation or header relationships, and distinguish genuine missing data from decorative blank space.

Worked example: Rename “Sheet3” to “Quarterly Donations,” label the amount column “Donation amount,” and widen it so values display fully. Users can identify both the worksheet's purpose and its data.

Mistake to avoid: Assuming that a cell's formula or value is usable because it exists internally.

7. Presentation titles and object sequence

Slide titles provide orientation and help users locate content. Give slides informative titles, and distinguish repeated topics when necessary. Arrange slide objects in a logical reading sequence using the authoring tool's ordering features. Visual placement and the sequence exposed to assistive technology can differ.

Worked example: Two slides become “Delivery options: collection” and “Delivery options: shipping.” On each, the title precedes explanatory text and the relevant diagram in the reading sequence.

Mistake to avoid: Assuming the object at the top of a slide will automatically be read first.

8. Alternative text based on purpose

Alternative text should convey the information or function an image contributes in its context. Identify what a reader would miss without the image, then express that meaning concisely. The same image can require different alternatives in different documents; appearance is relevant only when it supports the document's purpose.

Worked example: A photograph accompanies an inspection report. “Cracked seal around the north window” communicates the documented defect more usefully than “Photograph of a window.”

Mistake to avoid: Describing every visible detail while omitting the image's relevant message.

9. Text equivalents for complex visuals

Charts, diagrams, and other complex visuals may need a short alternative plus a longer description or accessible data equivalent. Preserve important relationships, trends, and values according to the purpose. Make the detailed equivalent easy to locate, and ensure it remains usable after conversion to the final format.

Worked example: A chart shows requests rising from 24 to 36. A nearby description states the increase of 12, or 50%, and an accessible table supplies the underlying values.

Mistake to avoid: Using “Chart of requests” as the only equivalent for a chart's findings.

10. Information beyond color

Color can reinforce meaning, but essential distinctions need another available cue. Use direct labels, symbols with explanations, patterns, or text as appropriate. The additional cue must identify the same information rather than merely adding decoration. This supports readers who cannot distinguish the chosen colors or perceive the visual display.

Worked example: A schedule marks available sessions in green and full sessions in red. Adding the words “Available” and “Full” preserves the status when color is unavailable.

Mistake to avoid: Changing red to another color without adding an independent way to identify status.

11. Text contrast calculations

Contrast ratio compares relative luminance using (lighter luminance + 0.05) divided by (darker luminance + 0.05). WCAG 2.1 AA generally requires 4.5:1 for ordinary text and 3:1 for qualifying large text, with defined exceptions. Evaluate the actual foreground and background combination, including backgrounds behind text.

Worked example: Luminances of 0.45 and 0.05 produce 0.50 divided by 0.10, or 5:1. That combination meets the ordinary-text contrast requirement.

Mistake to avoid: Estimating contrast from how bright or dark two colors appear to you.

12. Readable typography and Unicode text

Use actual text with correct character encoding rather than pictures of words or decorative substitutions that alter meaning. Unicode identifies characters independently of their visual design. Font choice, spacing, alignment, and size affect visual readability, while correct character mapping supports searching, copying, conversion, and assistive technology.

Worked example: A heading made from individual letter images is replaced with actual text. It can now be searched and copied, while a readable font preserves the intended visual emphasis.

Mistake to avoid: Assuming that a visually correct glyph necessarily has the correct underlying character.

13. Meaningful links and references

Link text should identify its purpose within the available context. Instructions and cross-references should identify content by meaningful names rather than relying solely on position, shape, or color. These choices support readers navigating nonvisually and make references more durable when layout changes.

Worked example: Replace “Click the blue link below” with “Read the equipment return instructions.” The destination and action remain understandable without seeing the link's color or position.

Mistake to avoid: Using several identical “More information” links when their destinations have different purposes.

14. Universal design and learning choices

Universal design seeks usability across diverse abilities through equitable, flexible, understandable, perceptible, forgiving, low-effort designs with appropriate space. Universal Design for Learning emphasizes options for engagement, representation, and action or expression. Apply these approaches without assuming that offering multiple formats excuses barriers in the primary document.

Worked example: A training document combines a structured explanation, a labeled diagram, and a text description. Readers can approach the same concept through different representations without losing essential information.

Mistake to avoid: Offering a visual alternative that is itself inaccessible to the intended learners.

15. Preserving accessibility during conversion

Accessible source structure supports conversion to PDF, EPUB, large print, and Braille, but conversion still needs verification. Use export methods intended to preserve document structure and accessibility information. Inspect the final deliverable because headings, reading order, image alternatives, and table relationships may change or disappear.

Worked example: A structured report exports to PDF with headings intact but one chart description missing. The final PDF needs that description restored before distribution.

Mistake to avoid: Assuming an accessible source guarantees an accessible result from every conversion method.

Remediating Electronic Documents

16. Choosing the repair format

Repair the authoring source when available and suitable, especially when the document will be revised or regenerated. Final-format remediation is necessary when the source is unavailable or cannot practically be corrected. Preserve repairs across the lifecycle by recording which changes must also be incorporated into future source versions.

Worked example: A quarterly report has incorrect headings in both its source and PDF. Correcting the source and exporting again prevents the same heading defect from returning next quarter.

Mistake to avoid: Repairing only the PDF while repeatedly publishing from an unchanged defective source.

17. Remediation dependencies and sequence

Plan repairs around dependencies. Image-only text needs recognition before reliable text tagging; structural corrections establish the context for lists, tables, and figures. Check reading order and descriptions after the content has been organized. This sequence reduces repeated work without removing the need to inspect the completed document.

Worked example: A scanned newsletter receives OCR first, then heading and paragraph structure, then a corrected column sequence. Descriptions are reviewed in their final positions.

Mistake to avoid: Carefully tagging an image-only page before establishing usable underlying text.

18. Captions and alternative text

A visible caption and an image alternative serve related but different access paths. A caption may identify a figure, give attribution, or summarize a point; alternative text conveys otherwise unavailable image meaning. Read both together and avoid needless repetition, while ensuring that essential visual information remains available.

Worked example: The caption “Figure B: assembly fault” identifies a photograph. Its alternative adds “The upper bracket is attached upside down,” supplying the information the caption omits.

Mistake to avoid: Copying a vague caption into alternative text and leaving the visual finding unexplained.

19. Decorative content and PDF artifacts

In PDF, mark purely decorative or repetitive non-content elements as artifacts so they do not interrupt meaningful reading. Determine their role in context before removing them from the logical structure. A visually minor mark may still communicate essential information, and meaningful running information needs an accessible representation.

Worked example: A repeated border is artifacted. A warning symbol remains meaningful content and receives an equivalent explaining the warning because removing it would lose information.

Mistake to avoid: Artifacting every image or every header without checking whether it conveys unique meaning.

20. Repairing PDF heading semantics

A PDF heading needs an appropriate heading tag, correctly associated content, and a place in the logical hierarchy. Inspect both the tag type and the text it contains. A tag named as a heading is insufficient when it contains an entire section or fails to correspond to the actual heading.

Worked example: A section title and three paragraphs are grouped inside one heading tag. Separate the title into the heading and place the paragraphs in paragraph tags beneath it.

Mistake to avoid: Changing tag names without inspecting which content each tag actually references.

21. Repairing PDF list structure

PDF list semantics distinguish the list container, list items, labels, and item bodies. Common standard tags include L, LI, Lbl, and LBody. Organize each bullet or number with its corresponding content, and place nested lists within the appropriate parent item so their relationships survive nonvisual navigation.

Worked example: A bullet and its sentence are separate paragraphs. Repair them as one list item with the bullet in Lbl and the sentence in LBody, inside L.

Mistake to avoid: Placing all bullets in one label tag and all sentences in another disconnected container.

22. Repairing complex table relationships

Table remediation requires accurate table, row, header, and data-cell semantics, plus the relationships needed to interpret each value. Spanning headers can require more explicit associations than simple tables. Simplify the table when possible, but preserve its information and verify how relevant reading software conveys the relationships.

Worked example: A sales table has a “West” header spanning “Online” and “Store.” The online value must be associated with both “West” and “Online,” not just the nearest visual label.

Mistake to avoid: Marking a header row while leaving multi-level header relationships unresolved.

23. Logical reading order

Reading order should preserve the author's intended meaning independently of visual arrangement. Multi-column layouts, callouts, footnotes, and floating images often expose errors. In PDF, inspect the logical structure and test actual reading behavior; changing an element's visible position does not necessarily change its accessible sequence.

Worked example: A two-column article alternates left and right lines during reading. Reorganize its logical sequence to finish the left column before continuing into the right column.

Mistake to avoid: Approving reading order after checking only the page's visual appearance.

24. Standard tags and role mapping

Standard PDF structure types communicate recognized semantic roles. Custom tags need appropriate role mapping so their intended meaning can be interpreted. Mapping must reflect the content's actual role; it cannot compensate for misplaced content or an incorrect hierarchy. EPUB instead relies on its applicable markup and semantic vocabulary.

Worked example: A publishing system creates a custom “SectionLabel” tag for genuine headings. Mapping it to the appropriate standard heading role preserves meaning, followed by a hierarchy check.

Mistake to avoid: Mapping every unfamiliar tag to a paragraph without examining its purpose.

25. Document titles and properties

Document metadata can provide a meaningful title and other identifying information distinct from a filename or visible cover heading. Populate the relevant properties and verify the title presented by the intended reader application. Properties should describe the actual document rather than retain template information.

Worked example: The file “report_final_7.pdf” has the metadata title “Community Garden Water Survey.” Updating its display settings where supported makes that meaningful title available to readers.

Mistake to avoid: Leaving the template's title “Untitled Project” in an otherwise remediated document.

26. Document and passage language

Set the document's primary natural language and identify substantive language changes where the format supports them. Language information helps speech technology select appropriate pronunciation. Use suitable language identifiers; add regional specificity when relevant and supported rather than choosing a locale merely because it is the application's default.

Worked example: An English report includes a Spanish sentence. English is set as the document language and Spanish for that sentence, allowing supported speech output to change pronunciation appropriately.

Mistake to avoid: Setting the whole document to Spanish to correct one Spanish quotation.

27. Recognizing scans and applying OCR

A page that looks like text may contain only an image. Investigate selection, search, extraction, and the underlying content rather than relying on appearance alone. OCR can create machine-readable text, but it does not automatically create headings, correct reading order, accurate tables, or a fully accessible document.

Worked example: A scanned letter cannot be searched for its subject line. After OCR, the subject becomes searchable; paragraph structure and reading sequence still require inspection and repair.

Mistake to avoid: Declaring a scanned document accessible immediately after OCR completes.

28. Verifying recognized text

Compare OCR output with the original, concentrating on consequential content such as names, amounts, dates, symbols, and instructions. Inspect word spacing and column boundaries as well as individual characters. Searchability or a tool's confidence indicator cannot establish that the recognized text accurately communicates the source.

Worked example: OCR reads an invoice total of 108.50 as 108.5O. Correct the final letter to zero and verify that copying and speech output expose the intended amount.

Mistake to avoid: Checking only ordinary prose while overlooking errors in financial or instructional details.

29. Character mapping and replacement text

When visible glyphs do not produce correct characters, repair the mapping where possible or use appropriate replacement text for the affected content. In PDF, ActualText can supply the intended textual representation. This differs from alternative text that explains an image's meaning; use the mechanism that matches the defect.

Worked example: A decorative ligature makes “office” extract as “o□ce.” Correct mapping or suitable replacement text restores “office,” which is then checked through copying and speech output.

Mistake to avoid: Adding an image description when the actual problem is incorrect text extraction.

30. EPUB structure and navigation

EPUB remediation involves the publication's content markup, reading sequence, navigation, language information, and text alternatives. Visual styling should agree with meaningful structure. Check both navigation destinations and the content they identify, because a usable table of contents cannot compensate for inaccessible or disordered sections.

Worked example: An EPUB contents entry labeled “Equipment care” opens the wrong chapter. Correct its destination and verify that the chapter begins with the matching semantic heading.

Mistake to avoid: Applying PDF tag-tree procedures directly to EPUB without considering its different structure.

Auditing and Testing Document Accessibility

31. Automated checks as an initial assessment

Begin evaluation with an appropriate automated checker to identify detectable defects efficiently. Use its results to guide repairs and manual inspection rather than as a complete verdict. Record the document version and checking context so later results can be compared meaningfully with the same deliverable.

Worked example: A checker flags a missing document language and two undescribed figures. Those findings are repaired, then a manual review examines whether the new descriptions actually convey the figures' meaning.

Mistake to avoid: Stopping evaluation when the automated checker reports no remaining errors.

32. Checker coverage and limitations

Different checking tools evaluate different rules and represent findings differently. A check may confirm an attribute's presence without judging its accuracy, while another tool may inspect additional structural constraints. Understand the selected tool's documented coverage and investigate disagreements rather than assuming one result invalidates all others.

Worked example: One checker accepts a figure because alternative text exists. Manual inspection finds the alternative says “image,” so the information barrier remains despite that structural pass.

Mistake to avoid: Treating identical-looking pass reports from different tools as equivalent evidence.

33. Manual semantic inspection

Compare the document's intended organization with its exposed structure. Inspect whether headings, paragraphs, lists, tables, and figures have the right roles and boundaries. A formally valid structure can still misrepresent the content, so use the visible document and its meaning as evidence during inspection.

Worked example: An entire page is tagged as a paragraph, including two section headings. Split the content into appropriate semantic elements so the sections become navigable.

Mistake to avoid: Accepting valid tag syntax as proof that the document's semantics are correct.

34. Choosing assistive technology tests

Select tests that expose different barriers. Screen readers convey structure and interaction through speech or Braille; magnification supports enlarged visual access; text-to-speech emphasizes spoken text; voice recognition can support command-based interaction. No single technology represents every user, and results depend on the document, reader application, and technology combination.

Worked example: A report is tested with a screen reader for heading navigation and magnification for crowded labels. Each test reveals a different aspect of usability.

Mistake to avoid: Assuming one successful assistive technology test establishes universal compatibility.

35. Screen reader navigation and traversal

Test both continuous reading and structural navigation. Continuous reading can reveal sequence and duplication problems; navigation by headings or links can expose missing structure and ambiguous labels. Compare what is announced with the intended content, including information that appears visually but is omitted from the accessible representation.

Worked example: A report reads smoothly from start to finish, but heading navigation finds only its title. The visually styled section labels need genuine heading semantics.

Mistake to avoid: Testing only continuous speech and never checking the document's navigation mechanisms.

36. Testing table interpretation

Navigate representative data cells with a screen reader and check whether the relevant row and column context is conveyed. Include edge cells, spanning headers, and transitions between sections. The goal is understandable relationships, not a particular spoken phrase that every technology must reproduce identically.

Worked example: In a donations table, a cell announces “250” without its campaign or month. Repair the header relationships and verify that the amount can be interpreted in context.

Mistake to avoid: Testing only the first data cell of a table with complex headers.

37. Text-to-speech and text integrity

Text-to-speech is useful for detecting recognition errors, missing spaces, and awkward text sequence. It does not necessarily expose the same structural information or interaction capabilities as a screen reader. Pair listening with text inspection, especially where pronunciation could conceal incorrect characters or altered punctuation.

Worked example: A scanned notice sounds like one continuous word because OCR omitted spaces. Correcting the recognized text restores separate words, then copying confirms the repair.

Mistake to avoid: Using acceptable speech output as the only evidence of correct document structure.

38. Magnification, zoom, and reflow

Inspect enlarged content for clipping, overlap, pixelation, and difficult navigation. Where the format and reader support reflow, check whether relationships and sequence survive it. Distinguish genuine two-dimensional content, such as some data tables, from ordinary paragraphs that should not depend unnecessarily on a rigid visual arrangement.

Worked example: At increased zoom, a worksheet truncates a long category label. Wrapping the text and adjusting the row height make the complete label visible.

Mistake to avoid: Approving enlarged text while ignoring information hidden outside its container.

39. Testing meaningful graphic contrast

Under WCAG 2.1 AA, graphical parts required to understand content generally need at least 3:1 contrast against adjacent colors, unless a particular presentation is essential. Identify which visual components carry meaning before measuring them. Decorative graphics are different from boundaries, lines, or symbols needed to interpret information.

Worked example: A necessary chart line has a measured contrast of 2.4:1 against its background. Changing its color produces 3.6:1, resolving that contrast defect.

Mistake to avoid: Measuring a chart's outer border while overlooking the low-contrast data lines.

40. Evaluating alternative text in context

Judge image alternatives against the surrounding content and the task the image supports. Check accuracy, essential meaning, unnecessary duplication, and access to any longer equivalent. An alternative's length is not a reliable quality measure: a short functional image and a complex explanatory diagram have different needs.

Worked example: A route diagram's alternative says “Map of the building.” It is revised to identify the accessible entrance and route, with detailed directions provided nearby.

Mistake to avoid: Approving alternative text solely because it is present or below a chosen word count.

41. Keyboard access and focus sequence

Test interactive elements without a pointing device. Check that users can reach controls, identify the current focus, activate functions, and leave components without becoming trapped. Focus order should support the task. Keyboard interaction and the document's continuous reading order are related but require separate verification.

Worked example: A PDF form's focus moves from address to submit before reaching the contact field. Correct the interaction sequence and verify that every field remains reachable.

Mistake to avoid: Assuming correct reading order automatically produces a usable tab sequence.

42. Form labels and instructions

Interactive fields need accessible names that identify their purposes, plus available instructions for formats, required information, or constraints. Check the name presented by assistive technology and its agreement with the visible label. A technical field identifier is not necessarily a useful user-facing label.

Worked example: A field announced as “Text17” is labeled “Preferred contact number.” Its accessible name is corrected, and a nearby instruction explains whether an extension can be included.

Mistake to avoid: Relying on placeholder text that disappears when users enter a value.

43. PDF link annotations and structure

A usable PDF link needs meaningful text, a correct destination, and an appropriate relationship between its interactive annotation and logical structure. Test both the announced purpose and activation. A phrase that looks linked can lack functioning interaction, while a functioning annotation can remain poorly represented to assistive technology.

Worked example: “View the venue directions” activates correctly with a mouse but is missing from structural link navigation. Repair the link's structural association and test it again.

Mistake to avoid: Checking only that clicking the blue underlined text opens a destination.

44. Embedded media accessibility

Evaluate embedded media as content and as an interaction. Speech and relevant sounds may need captions or transcripts; essential visual information may need description, depending on the media type and applicable requirements. Also inspect player controls and keyboard access. A text equivalent alone does not repair an inaccessible player.

Worked example: A training video shows an unspoken menu choice. Captions cover the narration, but description is added for that visual action and the player is tested by keyboard.

Mistake to avoid: Assuming accurate captions also convey every important visual event.

45. Actionable findings and regression checks

Document findings with a location, observed barrier, relevant requirement, reproduction method, and proposed repair. Retest the repaired feature and related content that the change could affect. Separate confirmed defects from observations needing investigation so authors and reviewers can make clear decisions.

Worked example: After repairing a table header, retesting confirms useful header announcements but finds that the last row was accidentally omitted. Restore that row and repeat the affected checks.

Mistake to avoid: Closing a finding because a repair was attempted rather than because its outcome was verified.

Document Accessibility Planning and Training

46. Lifecycle workflow and ownership

Build accessibility into intake, authoring, conversion, evaluation, remediation, approval, and publication. Assign responsibility and define the evidence needed at each handoff. Include later revisions so approved documents do not silently lose accessibility. Clear ownership prevents necessary tasks from falling between teams.

Worked example: An author prepares a structured source, a reviewer checks the exported file, and a publisher releases that reviewed version. A later revision returns through the same review step.

Mistake to avoid: Assigning accessibility to the final publisher after every earlier production decision is fixed.

47. Accessible templates and reusable components

Templates can provide suitable styles, layout patterns, table structures, and document properties before authors add content. Test them with realistic information and explain how to use their features. A template reduces repeated defects, but inserted images, altered styles, and unusual content still require review.

Worked example: A newsletter template includes heading styles and an image-description prompt. Testing with a long headline reveals clipping, which is corrected before the template is distributed.

Mistake to avoid: Treating template approval as automatic approval of every document produced from it.

48. Accessibility in automated document generation

Generated documents depend on both templates and variable data. Test conditional sections, empty values, long strings, multiple pages, and different data relationships. A sample with ordinary data may hide structural failures. Repair the generation pathway where possible so defects do not recur across large numbers of outputs.

Worked example: A statement generator produces correct one-page files but loses table headers on longer outputs. A multi-page test exposes the branch that needs correction.

Mistake to avoid: Validating only the default example supplied with the generation system.

49. Prioritizing document production and repair

Prioritize using barriers to essential tasks, audience needs, frequency of use, distribution, and opportunities to prevent recurring defects. Consider applicable obligations with the responsible organizational owners. Record why work is sequenced and how unmet access needs will be addressed while lasting repairs proceed.

Worked example: An inaccessible application form blocks submissions, while an old internal flyer has little ongoing use. The form receives immediate attention and its source template is repaired.

Mistake to avoid: Choosing the easiest files first while leaving a critical service document unusable.

50. Procurement and acceptance evidence

Specify the required deliverable formats, applicable accessibility criteria, source-file expectations, testing evidence, and responsibility for corrections. Evaluate representative outputs rather than relying solely on a vendor's general accessibility statement. Acceptance checks should cover the features and content the organization will actually distribute.

Worked example: A remediation supplier returns tagged PDFs and test records. Review includes a complex table and scanned appendix, revealing whether the agreed work covers those document features.

Mistake to avoid: Accepting “accessible PDF” as a complete specification without defining verification or correction responsibilities.

51. Role-based modular training

Organize training around the decisions each role makes. Authors need semantic creation skills, reviewers need testing skills, and publishers need reliable handoffs. Short, focused modules can be reused as references, but each should connect procedures to their accessibility purpose and demonstrate a meaningful result.

Worked example: Spreadsheet authors learn header labeling and readable data regions; publication staff learn to verify the exported file. Both modules use the same report to show the handoff.

Mistake to avoid: Giving every staff member identical tool instructions regardless of their production responsibilities.

52. Accessible instructional materials

Training must model the accessibility it teaches. Provide structured materials, clear step-by-step instructions, usable examples, and appropriate equivalents for images and multimedia. Explain visual actions in words. A learner should be able to understand the procedure without depending on an unexplained cursor movement or screenshot.

Worked example: A tutorial says “Apply the Heading 2 style to ‘Equipment care’” and provides an accessible practice document. Captions and spoken descriptions cover the demonstrated action.

Mistake to avoid: Teaching accessibility through an unlabeled screenshot sequence with no usable text instructions.

53. Reference guides and resource maintenance

Keep reference materials focused, searchable, and consistent with organizational procedures. Record their owner, applicability, and revision information. Vet external resources before recommending them, particularly where tool behavior changes. Archive superseded guidance so staff can identify the current procedure without losing useful historical context.

Worked example: An export guide is updated after a tool change. The old edition is clearly archived, and the current guide includes a final-file verification step.

Mistake to avoid: Leaving contradictory instructions equally prominent in the staff resource library.

54. Support and reproducible escalation

Collect enough information to reproduce a reported barrier: the document, affected task, application, assistive technology, observed result, and relevant settings. Determine whether the cause is content, conversion, generation, or software interaction. Give staff actionable support and escalate documented software problems to the appropriate vendor or technical owner.

Worked example: A user reports missing table headers. Support reproduces the problem in one reader application but not another and sends the vendor a minimal document demonstrating it.

Mistake to avoid: Blaming assistive technology before examining the document and reproducing the reported behavior.

Policy Promotion, Advocacy, and Advising

55. Different roles of accessibility standards

WCAG provides accessibility requirements; PDF/UA-1 addresses accessible use of PDF structure and features; the PDF specification defines the underlying format. The Matterhorn Protocol supports PDF/UA assessment through identified failure conditions. These resources complement one another, but passing one evaluation does not establish every other requirement or usability outcome.

Worked example: A PDF has a sound tag structure but unreadable text contrast. Structural conformance evidence does not resolve the separate visual accessibility barrier.

Mistake to avoid: Treating “valid PDF,” “PDF/UA conforming,” and “accessible for every task” as interchangeable claims.

56. Policy, procedure, and implementation plan

A policy states organizational commitments and expectations. A procedure explains how staff carry out those expectations. An implementation plan identifies work, responsibilities, resources, and sequencing. Keep them aligned: a strong commitment needs practical production steps, while a detailed procedure needs an agreed purpose and scope.

Worked example: A policy requires accessible public reports, a procedure defines source and export checks, and a plan assigns owners to replace defective report templates.

Mistake to avoid: Publishing a policy statement without establishing a workable process for fulfilling it.

57. Determining applicable obligations

Distinguish technical guidance from legal, contractual, and organizational obligations. Applicability can depend on jurisdiction, organization, service, audience, and agreement. Consult current authoritative sources and the responsible legal or policy owner instead of transferring requirements from another setting. Document the selected standards and their relevant editions.

Worked example: An international organization distributes reports through several regional services. Its policy owner determines applicable obligations before the document team selects a shared production baseline.

Mistake to avoid: Assuming a requirement cited in one country's guidance automatically governs every organization.

58. Advocacy through demonstrated barriers

Explain accessibility through specific participation barriers and concrete improvements. Demonstrations can connect semantic structure, keyboard access, or text equivalents to organizational tasks. Use respectful examples and evidence rather than portraying all disabled users as having identical needs or relying on unsupported financial promises.

Worked example: A demonstration compares navigating a long report with and without headings. The structured version lets a reader locate the expenses section directly, showing the practical value of semantic authoring.

Mistake to avoid: Using a single user's experience as a complete representation of every disability.

59. Reviewing policy and technical changes

Review policies and procedures when standards, tools, workflows, contracts, or user evidence change, as well as through planned organizational reviews. Assess the change, update affected guidance, and communicate the decision. A newer tool or standard should trigger evaluation rather than an unsupported assumption that existing documents are automatically compliant.

Worked example: A new export tool changes how image descriptions are preserved. The team tests representative outputs and updates the approved workflow after confirming the resulting differences.

Mistake to avoid: Replacing established guidance with new advice before checking its applicability and consequences.

60. Access requests and unresolved barriers

A document accessibility policy should provide a usable route for reporting barriers and requesting necessary formats. Identify who responds, how the required information will be made available, and how recurring problems reach production owners. Address immediate access needs while pursuing durable repairs; an alternative must preserve relevant information and function.

Worked example: A reader cannot use a complex report PDF. The organization supplies an accessible structured version and records the defect so the report-generation template is corrected.

Mistake to avoid: Providing an alternative that omits the data or task the reader actually needs.

Sources

Sources checked: Oct 2, 2026

Browse all study guides