Sample University, Disability Services Letterhead placeholder, uploaded once during account setup

Document Accessibility Check, Built for Non-Experts

Accessibility Audit Report

A compliance-informed audit of a publisher document. Findings written so they can be quoted directly in correspondence and procurement records.

Document title	Introduction to Statistics, Chapter 4: Probability Distributions
ISBN / DOI	978-0-XX-XXXXXX-X
Publisher	Sample Publisher Limited
Pages	42
Audit date	27 April 2026
Tracking ID	T6-2026-0427-A14F
Tool	EquitableDocs Tool 6, v0.1 preview

Does not conform

This document fails accessibility conformance against the standards listed below.

WCAG 2.2 Level AAOctober 2023Fails
PDF/UA-1ISO 14289-1:2014, Matterhorn Protocol v1.1Fails
EN 301 549 V3.2.1, Clause 10European harmonised standardFails
RPwD Act 2016, GIGW 3.0India accessibility obligationsPartial
UN CRPD Article 24Right to inclusive educationFails

Headline finding, quotable

Students with blindness and low vision do not have access to this document. The publisher has not added image descriptions to 12 images, has not set the correct reading order in 4 sections, and has not labelled the column headers in 1 table. Three of the sixteen issues prevent these students from using the document at all.

From the EquitableDocs Tool 6 Document Accessibility Check, 27 April 2026, sample document.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Executive summary

A faculty-readable overview, with quotable lines pulled out for correspondence.

1. Severity at a glance

Blocker	Prevents any meaningful access by screen reader users.	3
High-impact	Renders significant content unreadable or out of order.	8
Polish	Affects cleaner reading; not blocking.	5

2. Scope of this audit

EquitableDocs is the bridge and the last-mile delivery. This section names what the automated tool covers, what the EquitableDocs Accessibility Collective handles in the same workflow, and what is on the roadmap for upcoming versions.

What this audit covered, automated

Tool 6 v0.1 evaluated the document against seven categories of accessibility defect. Findings are evidence-based, with a screen-reader transcript and tag-tree screenshot for every defect entry.

Tag tree structure and reading order
Alt text presence and quality
Table header structure
Heading hierarchy and skipped levels
Form field labels
Document language declaration
COGA, plain-language and clutter heuristics

Last-mile remediation by EquitableDocs

No automated tool can reliably verify the items below. EquitableDocs handles them as part of the same workflow, through the Accessibility Collective: trained members with the relevant specialism complete the remediation. The institution does not need to find another vendor.

STEM equations and MathML markup. Routed to a Collective member with STEM specialism.
Complex diagrams (medical, engineering, design). Routed to a Collective member with the relevant domain specialism.
Scanned PDFs without a text layer. OCR conversion plus tag-tree remediation, completed by a Collective member.
Alt-text meaningfulness on figures. AI vision pass coming in v2; until then, every figure is reviewed by a Collective member.
Indic-script OCR depth. Reviewed by a Collective member with the relevant language.

Detection coverage, v2 roadmap

Coming in the next version of Tool 6.

PowerPoint detector module. .pptx files audited end to end.
Word inline-image vision pass. AI check on alt-text meaningfulness for Word.
Equation alt-text quality check. Pairing with Mathpix or equivalent.

Track B, v3 roadmap

Longer-horizon work for institutional partners.

Self-hosted kit. Universities run Tool 6 inside their own infrastructure.
Cross-institutional Publisher Accessibility Index. Aggregate publisher conformance scoring.

Boundaries of our claims

What this audit does not assert, by design.

Auto-tagging accuracy percentages are not published.
CommonLook table parity is not benchmarked.
The audit verdict is a reasoned outcome of running the named detectors, not a legal certificate.

3. Acceptance checklist

The document is ready for screen-reader users when every row below shows MET. Each row gives the statement in plain language, with the corresponding PDF / remediator term shown in muted text below.

Met

The document tells the screen reader what language it is in.

Document Lang attribute set at root. Matterhorn 11-001, PDF/UA-1 7.21.
Met

Headings appear in the right order, no levels are skipped.

Heading hierarchy: no H1 followed directly by H3. Matterhorn 14-002.
Met

Every form field has a label the user can see.

Visible labels on form fields. WCAG 2.2 AA SC 3.3.2; not placeholder-only.
Met

No tagged image carries a blank description in place of a real one.

Figure tags do not carry empty Alt where the figure conveys meaning.
Met

Form fields use a real label, not just a hint inside the box that disappears as you type.

No placeholder-only labels.
Not met

Every picture that means something has a description for screen readers.

Figures need descriptive Alt text. Matterhorn 13-002, 13-004; PDF/UA-1 7.3.
Not met

Every data table has labelled column headers so the screen reader can read the table cell by cell.

Header rows tagged with TH and Scope=Col (or Scope=Row). Matterhorn 15-002; PDF/UA-1 7.5.
Not met

The screen reader reads the page in the same order a sighted reader sees.

Tag-tree reading order matches visual layout. Matterhorn 09-001; PDF/UA-1 7.2.
Not met

Author lists and similar short-line groups are tagged as a list, not as separate paragraphs.

Author block uses L / LI / LBody. Matterhorn 09-002.
Not met

No empty space announces "blank" to a screen reader.

No Span ActualText="blank" remains. Matterhorn 04-001; PDF/UA-1 7.18.
Not met

When you re-run the Acrobat Accessibility Checker, no new failures appear.

Acrobat Pro Accessibility Checker, all checkpoints Passed.
Not met

When you run veraPDF, the document passes with no PDF/UA-1 errors.

veraPDF, zero PDF/UA-1 violations.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

4. What screen reader users encounter on each page

Each entry leads with the screen-reader experience in plain language. Technical detail (detector output, Matterhorn checkpoint, why-list, verify steps) sits in a single muted footer at the bottom of the entry, available for procurement and remediator records.

4.1 Tagging

D-001 High Page 1, structure position 17. Detector class: tagging.

What this sounds like

Each author name reads as a separate paragraph, with a half-second pause between names.

A blind student hears nine isolated names with audible silences between them, the same way a sighted reader would experience the page if every author name had a blank line above and below it. The block stops reading like a list and starts reading like nine unrelated facts.

Screen reader output

M Boutros [paragraph] G Awad [paragraph] A Adio [paragraph] [seven more, each as a paragraph]

Tag tree evidence

Fix in Acrobat Pro

Open Tags panel. Locate position 4 in the structure tree, page 1.
Select positions 4 through 12 (the nine  tags from "M Boutros" to the last author).
Right-click on position 4. New Tag, type List, title "Authors". Drag positions 5 to 12 inside the new <L>.
For each child, change tag from  to <LI>. Add an <LBody> inside.
Save. Re-run Acrobat Read Out Loud on page 1 to verify no paragraph pauses between names.

Every occurrence, ticked off as you fix

Nine occurrences. Work through each in order. Tick the box once the tag is corrected and saved.

	Page	Structure position	Current tag	Author name
☐	1	4	<P>	M Boutros
☐	1	5	<P>	G Awad
☐	1	6	<P>	A Adio
☐	1	7	<P>	R Sharma
☐	1	8	<P>	S Patel
☐	1	9	<P>	K Iyer
☐	1	10	<P>	P Mehta
☐	1	11	<P>	D Singh
☐	1	12	<P>	B Rao

Detector flagged: Nine consecutive tags each contain only a single author name. Author lists should not be tagged as paragraph runs.
Standards: Matterhorn 09-002WCAG 2.2 AA SC 1.3.1EN 301 549 10.1.3.1
Why it matters: Author list cannot be navigated as a single block; tag tree harder to audit; search-by-author may fail.
Acceptable patterns: Option A: <L> with nine <LI> children. Option B: one  with comma-separated names.
Verify with: Tag tree shows one <L> with nine <LI> tags; Acrobat Read Out Loud announces names without paragraph pauses; Acrobat Accessibility Checker passes.

D-002 Blocker Pages 4, 7, 12. Detector class: tagging. Twenty-eight occurrences.

What this sounds like

The screen reader says “blank” twenty-eight times across the document.

The student cannot tell whether each “blank” is a fill-in slot they need to write into, a layout gap they should skip past, or a genuine missing word in the source text. Comprehension breaks every few seconds; the student must re-read surrounding sentences to recover meaning.

Screen reader output

What is your guess? blank Subtract to test your answer. blank minus blank equal to blank Nate has blank shells left.

Tag tree evidence

Fix in Acrobat Pro

Open Tags panel. Use Find (Ctrl+F in the panel) for ActualText="blank".
Work through the occurrences in the table below. For each, page reference and position are given.
Right-click the Span at the named position. Properties > Tag tab > clear the ActualText value.
For occurrences that mark a fill-in slot, set ActualText="" and tag the slot as a form field.
Save. Re-run veraPDF on the file. Matterhorn 04-001 should report zero failures.

Every occurrence, ticked off as you fix

Twenty-eight occurrences across pages 4, 7, and 12. First six shown. Full enumeration in the attached defect log (Appendix H).

	Page	Position	Span content	Likely intent
☐	4	23	[blank]	fill-in answer slot
☐	4	24	[blank]	fill-in answer slot
☐	4	25	[blank]	fill-in answer slot
☐	4	47	[blank]	layout artefact
☐	7	18	[blank]	fill-in answer slot
☐	7	19	[blank]	fill-in answer slot

22 further occurrences listed in Appendix H. Defect log table.

Detector flagged: ActualText is set to "blank" on empty elements. Remove the ActualText or replace with an empty string so the screen reader does not announce "blank".
Standards: Matterhorn 04-001PDF/UA-1 7.18WCAG 2.2 AA SC 1.1.1EN 301 549 10.1.1.1
Why it matters: Reading flow breaks every few seconds; users cannot distinguish intentional from defective blanks; document fails PDF/UA-1 conformance.
Acceptable patterns: Remove ActualText entirely (Span becomes silent), or set ActualText="" and tag intentional blanks as form fields.
Verify with: Tag tree returns zero ; NVDA reads pages without "blank" announcements; veraPDF reports no Matterhorn 04-001 failures.

4.2 Reading order

D-003 Blocker Page 9, structure positions 4 to 7. Detector class: reading_order.

What this sounds like

On page 9, the answer is announced before the question.

A blind student hears the answer “twenty-five” first and the question “what is five squared” second. A sighted student would see them in the right order on the page; the blind student hears them backwards and cannot follow the lesson’s logic. The student has no way to know that the underlying file disagrees with the visible page.

Screen reader output

The answer is twenty-five. [paragraph] What is five squared? [paragraph]

Page canvas evidence

Fix in Acrobat Pro

Tags panel, page 9. Locate position 4 (currently the answer paragraph) and position 7 (currently the question).
Drag position 7 above position 4 in the structure tree.
The visible page layout must not shift; only the tag-tree order changes.
Save. Run Acrobat Read Aloud on page 9. The question must be announced before the answer.

Every occurrence, ticked off as you fix

One occurrence. The pattern may recur in other question-answer worksheets in the document; check pages with similar layouts after this fix.

	Page	Structure positions	Issue
☐	9	4 to 7	Answer paragraph at position 4 precedes question at position 7. Visual order has them reversed.

Detector flagged: Tagged reading order does not match visual order. Question text is at structure index 7; the answer is at index 4. Visual layout has the question above the answer.
Standards: Matterhorn 09-001PDF/UA-1 7.2WCAG 2.2 AA SC 1.3.2EN 301 549 10.1.3.2
Why it matters: Answer announced before question; student cannot follow without sighted help; pattern likely repeats in question-answer worksheets.
Acceptable patterns: Question paragraph must precede the answer paragraph or fill-in slot in tag-tree order.
Verify with: Acrobat Read Aloud announces question first; NVDA / VoiceOver re-test confirms order.

Thirteen further entries follow this same pattern across alt text, headings, tables, forms, and cognitive accessibility. Truncated in this preview.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Appendix A. Standards reference

Cite-able mapping for legal, compliance, and procurement correspondence.

Issue type	Matterhorn / PDF/UA-1	WCAG 2.2 AA	EN 301 549
Empty Span ActualText announces "blank"	Matterhorn 04-001 PDF/UA-1 7.18	SC 1.1.1	10.1.1.1
Reading order mismatch with visual layout	Matterhorn 09-001 PDF/UA-1 7.2	SC 1.3.2	10.1.3.2
Author block tagged as separate paragraphs	Matterhorn 09-002	SC 1.3.1	10.1.3.1
Image without alt text	Matterhorn 13-004 PDF/UA-1 7.3	SC 1.1.1	10.1.1.1
Table header row not marked TH with Scope	Matterhorn 15-002 PDF/UA-1 7.5	SC 1.3.1	10.1.3.1
Heading level skipped	Matterhorn 14-001 PDF/UA-1 7.4	SC 1.3.1	10.1.3.1
Document language not declared	Matterhorn 11-001 PDF/UA-1 7.21	SC 3.1.1	10.3.1.1

Standards versions applied. WCAG 2.2 AA, October 2023. PDF/UA-1, ISO 14289-1:2014. EN 301 549 V3.2.1 (2021-03), clause 10. Matterhorn Protocol v1.1, PDF Association, 2021.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Appendix B. Aggregate metrics

Document-level numbers for procurement records and aggregate reporting.

Total pages42PDF, version 1.7

Total issues16Across 7 detector classes

Pages with blockers37 percent of pages

Reading-time impact28 minEstimated overhead vs accessible baseline

Sighted-help moments3Where the student must ask for help

Acceptance checklist5 of 12Statements currently met

Detector classes that ran

Tagging, structure tree integrity, ActualText, role-mapping
Alt text presence and quality on figures and images
Reading order, comparing tag-tree order against visual layout
Tables, header row, scope attributes, layout-vs-data classification
Headings, hierarchy, skipped levels
Forms, label presence, accessible name
Language declaration at document and span level
COGA, plain-language and clutter heuristics

Reading-time impact, methodology

Reading-time overhead is estimated as the additional time a screen reader user takes to navigate the document compared with an accessible baseline. The estimate combines counts of "blank" announcements, reorder corrections, alt-text rereads, and table-navigation falls-back to default cell reading. The number is indicative, not measured per user.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Appendix C. Summary checklist before resubmission

Re-upload the file once each statement is true. Tool 6 will re-check and update the acceptance list.

The author block is tagged as a single  with comma-separated names, or as an <L> with nine <LI> children.
No  with ActualText="blank" remains. NVDA does not announce "blank" between content.
Every question paragraph appears in the tag tree before its answer paragraph or fill-in slot.
Every <Figure> has descriptive alt text. No placeholder alt text such as "image" or "graphic".
Every data table has a header row tagged with <TH Scope="col"> (or "row" as appropriate). No data table is tagged as a layout table.
The Acrobat Accessibility Checker reruns with zero new failures.
veraPDF passes with zero PDF/UA-1 violations.
Read Out Loud on two random pages flows naturally, with no strange pauses or repeated words.

Tools to re-run before resubmission

Acrobat Pro Accessibility Checker, all checkpoints Passed
veraPDF, zero PDF/UA-1 violations
NVDA on Windows or VoiceOver on Mac, two random pages spot-checked
Optional: PAC 2024 by axes4 for an independent PDF/UA-1 verdict

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Appendix D. Scope and limitations

Honest scope is the EquitableDocs voice signature. Read this before quoting numbers from the report.

What this report does claim

The named detector classes ran on the document and produced the findings shown.
The standards mappings in Appendix A are correct as of the standards versions cited.
The fix steps in each entry are correct for the named tool version.
The conformance verdict on the cover page is the result of running the listed detectors. It is a reasoned verdict, not a legal certificate.

What this report does NOT claim

This is not a legal certificate of WCAG 2.2 AA conformance.
The report does not measure auto-tagging accuracy.
The report does not measure CommonLook table parity, full STEM or MathML support, or full Indic-script OCR fidelity.
The reading-time impact metric is indicative, not measured against any specific reader.
The report does not benchmark against competing tools.

Tool version and roadmap

Tool 6 v0.1, the preview release. Deferred to later versions:

PowerPoint detector module, Tool 6 v2.
Word vision pass on inline images, v2.
Track B self-hosted kit, v3.

How to escalate

If the publisher does not respond within sixty days, the institution may escalate to: the Office of the Chief Commissioner for Persons with Disabilities, the UGC accessibility cell, or the regional disability commissioner.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Appendix E. Statements you can copy

Lines from this audit, written so they stand on their own. Grouped by who you are speaking to. Copy any of them as is.

If you only use one line, use this one

Students with blindness and low vision do not have access to this document. The publisher has not added image descriptions to 12 images, has not set the correct reading order in 4 sections, and has not labelled the column headers in 1 table.

From the cover page, headline finding.

Writing to the publisher, procurement, or legal Compliance-frame statements suitable for formal correspondence and procurement records.

This document does not meet WCAG 2.2 AA. It fails PDF/UA-1 against 7 of 31 evaluated Matterhorn checkpoints. Source: cover page, conformance verdict.
Tool 6 evaluated 31 PDF/UA-1 checkpoints against this document. Seven failed; twenty-four passed. PDF/UA-1 cannot be claimed. Source: compliance summary, this page.

Describing what the student experiences For faculty briefings, disability-services conversations, and slide decks. Plain language, reader-centred.

The screen reader says “blank” twenty-eight times across the document. The student cannot tell which is an answer slot and which is a defect. Source: defect D-002, page 4 of this report.
On page 9, the answer is announced before the question. A blind student hears them backwards and cannot follow the lesson’s logic. Source: defect D-003, page 4 of this report.
Each author name reads as a separate paragraph, with a half-second pause between names. The block stops reading like a list. Source: defect D-001, page 4 of this report.

Giving numerical context For grant proposals, annual reports, and procurement business cases.

Three of the sixteen issues found in this document require sighted assistance to work around. Source: aggregate metrics, page 5 of this report.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Appendix F. Compliance summary, all 31 evaluated checkpoints

PAC-style summary of which checkpoints fail and which pass. Cite-able for legal, compliance, and procurement correspondence.

Standard / checkpoint	Status	Failures	Plain-language note
Matterhorn 04-001 (PDF/UA-1 7.18), Span ActualText	Fail	28	Empty Span elements announce "blank" to screen readers.
Matterhorn 09-001 (PDF/UA-1 7.2), reading order	Fail	4	Some paragraphs are tagged out of visual order.
Matterhorn 09-002, paragraph fragmentation	Fail	9	Author block tagged as separate paragraphs, not a list.
Matterhorn 13-002, figure alt-text quality	Fail	3	Three figures use placeholder alt text such as "image".
Matterhorn 13-004 (PDF/UA-1 7.3), figures present	Fail	12	Twelve figures have no alt text attribute at all.
Matterhorn 15-002 (PDF/UA-1 7.5), table headers	Fail	1	One table missing TH and Scope.
Matterhorn 28-002, form-field tooltips	Fail	2	Two form fields have no tooltip / accessible name.
Matterhorn 01-001, real content tagged	Pass	0	All real content is reachable through the structure tree.
Matterhorn 01-003, structure type nesting	Pass	0	Nested types follow the PDF/UA-1 nesting rules.
Matterhorn 02-001, role mapping to standard types	Pass	0	Custom tags are role-mapped to standard structure types.
Matterhorn 06-001, document title	Pass	0	Document title is set in metadata.
Matterhorn 06-002, ViewerPreferences DisplayDocTitle	Pass	0	The reader displays the document title, not the file name.
Matterhorn 07-001, suspect markings	Pass	0	No suspect-content markers remain in the structure tree.
Matterhorn 09-003, sect tag use	Pass	0	Sect tags wrap genuine document divisions, not single headings.
Matterhorn 09-004, block-level structure	Pass	0	Block-level content is tagged with block-level structure types.
Matterhorn 10-001, role mapping completeness	Pass	0	Every custom role maps to a defined standard type.
Matterhorn 11-001 (PDF/UA-1 7.21), document language	Pass	0	Language declared as en-US at the Document level.
Matterhorn 11-002, span language attributes	Pass	0	No language changes mid-paragraph require a Span Lang.
Matterhorn 14-001 (PDF/UA-1 7.4), heading levels exist	Pass	0	Headings are tagged using H1 to H6 elements.
Matterhorn 14-002, heading hierarchy not skipped	Pass	0	No heading level is skipped (no H1 followed directly by H3).
Matterhorn 15-001, table summary attribute	Pass	0	Data tables include a Summary attribute where required.
Matterhorn 15-003, layout vs data tables	Pass	0	Layout tables are marked as Artifact, not as data tables.
Matterhorn 17-001, list structure	Pass	0	Lists use L, LI, and LBody where present.
Matterhorn 17-002, list labels	Pass	0	List item labels (Lbl) are present where the list shows numbers or bullets.
Matterhorn 19-003, optional content	Pass	0	Optional content is tagged so it is reachable when activated.
Matterhorn 21-001, native PDF objects, page boundaries	Pass	0	Page-break artifacts do not appear inside content tags.
Matterhorn 25-001, annotations tagged	Pass	0	Hyperlinks and other annotations are tagged with Link wrappers.
Matterhorn 26-001, action triggers	Pass	0	No action triggers depend on visual or pointer input only.
Matterhorn 28-001 (PDF/UA-1 7.18), form fields tagged	Pass	0	All form fields are tagged in the structure tree.
WCAG 2.2 AA SC 1.4.3, contrast minimum	Pass	0	Text contrast meets the 4.5 to 1 minimum on all spot-checked pages.
WCAG 2.2 AA SC 3.3.2, visible form labels	Pass	0	All form fields have visible labels, not placeholder-only labels.

Summary: 7 of 31 evaluated checkpoints fail; 24 pass. PDF/UA-1 cannot be claimed for this document.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Appendix G. Full screen-reader transcript

A literal record of what NVDA reads aloud across the document. Lines from this transcript appear as evidence in defect entries; the appendix is the complete record. Use this as an evidence artefact for re-audit after remediation, or as supporting material in correspondence with the publisher.

Page 1 heading level 1: Introduction to Statistics heading level 2: Chapter 4: Probability Distributions M Boutros [paragraph] G Awad [paragraph] A Adio [paragraph] R Sharma [paragraph] S Patel [four more authors as paragraphs] Department of Mathematics Page 2 heading level 2: Abstract This chapter introduces probability distributions used in undergraduate statistics. graphic heading level 3: Learning objectives Page 4 heading level 2: Practice questions Question 1: Nate finds 50 shells. He gives 20 shells to Josh. How many shells does Nate have left? What is your guess? blank Subtract to test your answer. blank minus blank equal to blank Nate has blank shells left. Page 7 Question 2: Anna needs 70 beads for a craft. She has 30 beads. How many more beads does Anna need? Anna needs blank more beads. graphic Page 9 heading level 3: Worked example The answer is twenty-five. [paragraph] What is five squared? [paragraph] Solution method. Page 12 Question 3: Lucy reads for 20 minutes. Ty reads for 10 minutes less than Lucy. How many minutes does Ty read? Ty reads for blank minutes. Page 15 heading level 2: Summary A probability distribution describes the likelihood of each possible outcome of a random variable. graphic End of chapter.

Page-by-page captions: orange page-marker headers indicate page number transitions. Amber-highlighted text marks where a defect was detected. The "Hear it aloud" button reads the full transcript using the browser's default voice. This is a simulation of how a screen reader announces the document; it is not the actual NVDA voice. For full fidelity, run NVDA on the source PDF.

Sample University, Disability Services Tracking ID T6-2026-0427-A14F

Appendix H. Defect log, all sixteen entries

CSV-equivalent log of every defect found in the document. One row per defect. Use this as a project-management tracker through remediation, or import to a spreadsheet for procurement records. A machine-readable .csv version is shipped alongside this report.

ID	Severity	Page	Position	Detector class	Headline, plain language
D-001	High	1	4 to 12	tagging	Each author name reads as a separate paragraph
D-002	Blocker	4, 7, 12	multiple	tagging	Screen reader announces "blank" twenty-eight times
D-003	Blocker	9	4 to 7	reading_order	Answer announced before the question
D-004	High	2	3	alt_text	Title page logo has no description
D-005	Polish	3	12	alt_text	Decorative divider should be tagged as artifact
D-006	Blocker	5	7	alt_text	Probability tree (Figure 1.1) has no description
D-007	High	6	4	table_headers	Table missing TH and Scope on header row
D-008	High	7	22	alt_text	Bar chart (Figure 1.2) has no description
D-009	High	8	14	alt_text	Venn diagram (Figure 1.3) has no description
D-010	High	10	9	alt_text	Histogram (Figure 1.4) has no description
D-011	High	11	17	alt_text	Scatter plot (Figure 1.5) has no description
D-012	High	11	28	alt_text	Regression line (Figure 1.6) has no description
D-013	Polish	13	5 to 6	reading_order	Footnote read before main paragraph it qualifies
D-014	Polish	14	11	reading_order	Page number read mid-sentence
D-015	Polish	15	8	alt_text	Equation graphic without alt text
D-016	Polish	16	3	alt_text	Decorative bullet treated as content

Severity counts: 3 blockers, 8 high-impact, 5 polish. Total 16 defects across 7 detector classes. The .csv companion file uses the same column order and is suitable for import into Excel, Google Sheets, or any project management tool.