Creating and Fixing Accessible PDFs

This guide covers how to work with PDFs so they meet accessibility standards, including how to improve existing PDFs and how to choose a better format when a PDF isn't necessary.

PDFs are one of the most common files instructors upload, and one of the most inaccessible. Scanned readings, untagged exports from Word or Google Docs, and older course files often score lowest on accessibility checks because assistive technology cannot recognize their structure, text, or images. Many PDFs come from scanning paper documents, which produces image based PDFs. These are completely inaccessible until the text is converted using Optical Character Recognition (OCR) software.

Consider an accessible alternative first

Use a Canvas page (HTML) instead of a PDF For syllabus information or assignment instructions, copy the text directly into a Canvas page rather than uploading a PDF.

Note: if you must scan a document, avoid creating an image only PDF. When text is locked inside an image, screen readers cannot access it. Use a scanner that offers OCR, which converts image based text into readable, searchable text and dramatically improves accessibility.

Accessibility can't be judged just by looking at a PDF. PDFs need a structure called tagging so screen readers can navigate them properly. For example, text might look like a heading because it's larger or bold, but if it isn't tagged as a heading, a screen reader won't recognize it as one. Untagged PDFs cause real accessibility problems, and scanned PDFs are worse still, since scanned pages are read as images rather than text and are completely inaccessible.

Start with the original document If you created the material, the most accessible version is usually your original file, not the exported PDF.

Check for library hosted versions Before scanning a reading, check your library or the publisher's website. Many journal articles and book chapters are already available online in accessible formats, and linking to the official version usually avoids the need to scan at all.

How to check if a PDF is accessible

• Try selecting text. If you can't select individual text and instead select the whole page, it's likely an image based PDF and is inaccessible.
• Check your accessibility indicators in Canvas. Use your course's accessibility report to find the lowest scoring files. Look for PDFs with red or orange scores, then click the score to see the specific issue. Common messages include:
  • "This PDF is scanned," meaning the document is image based and the text can't be read or navigated by screen readers.
  • "This PDF is untagged," meaning the document lacks the tag structure that lets assistive technology navigate headings, lists, tables, and reading order.

Create accessible PDFs from the start

The most reliable way to get an accessible PDF is to start from an accessible source document. Features like headings, alt text, and reading order need to be built into the original file (Word, Google Docs, PowerPoint, Google Slides) before you convert it to PDF. Skip this step and the resulting PDF will likely be missing the tags it needs.

A well built source file should include:

• Proper heading structure using built in styles (Heading 1, Heading 2, and so on)
• Meaningful alt text for all images, charts, and graphics
• Clear, descriptive hyperlinks, not "click here"
• Legible fonts, good color contrast, and clean formatting with no text boxes or floating elements
• Simple tables with headers, avoiding merged or nested cells
• A built in accessibility check run before exporting

Keep the native format when you can. Before converting to PDF, consider keeping the document in its native Google or Microsoft format. Uploading these directly to Canvas lets the built in previewer display the content accessibly and makes future edits easier.

Export with the right settings.

• Microsoft Word or PowerPoint: File, Save As, PDF. Make sure "Document structure tags for accessibility" is checked in the options.
• Google Docs or Slides: Native exports keep much of the structure but may lack full tagging. Use the Grackle plugin to scan for issues and export a properly tagged PDF. After fixing the errors Grackle finds, use its dedicated "Export to PDF" button rather than the standard File, Download menu, since only that button produces a fully tagged PDF.
• Never "print to PDF." Printing flattens the document into an image and destroys all accessibility tags. Always use Save or Export.

Remediate an existing PDF

If you don't have the source file, or you have no better option than a PDF, here's how to fix one that already exists.

For untagged PDFs or image based scans: start with automated prep, then do a manual review.

Step 1: Automated prep Use a conversion tool such as SensusAccess to do the first pass on image based scans or untagged documents. Results vary depending on the quality of the source file, and converted files usually still need manual fixes like headings and alt text.

Step 2: Manual review using Ally Use this step after automated prep, or for PDFs that are already tagged but need alt text or further cleanup.

1. Check Ally's report for the file to see the specific issues it flags, such as missing tags, a missing document title, a missing language setting, or missing alt text on images.
2. Fix each issue Ally identifies, adding a document title, setting the language, and adding alt text as needed.
3. Re-check the file's Ally score to confirm the issues have been resolved.
4. Save the updated file and upload it to Canvas.

Pro tip: an AI tool can help speed up the process of drafting initial image descriptions for you to review and refine.

Math and STEM content in PDFs

Math introduces its own accessibility challenges, especially in exported or scanned PDFs. Equations and notation often end up as images, which screen readers can't interpret. Even a tagged PDF can struggle with complex math. Your approach depends on whether you have access to the original source file.

If you have the original source file

Author math in a format that supports real, accessible math rather than images:

• Canvas Rich Content Editor: Canvas supports LaTeX and MathML, which are more accessible than images of equations. Publish equations directly in Canvas instead of embedding them in a PDF. Open the Math Editor from the editor's Insert menu and choose Equation. It handles everything from simple equations to complex upper level or graduate notation.
• Produce HTML from LaTeX: For documents written in LaTeX, HTML output is often easier to make accessible than PDF output. Reference: LaTeX markup best practices for accessible HTML.
• Produce accessible PDFs from LaTeX Beamer slides: the standard Beamer class has historically struggled to produce accessible PDF slides, since tagging and reading order are often missing. The ltx-talk class is a newer option that can produce accessible slide presentations while keeping the frame syntax people are used to from Beamer. It requires switching to the LuaLaTeX compiler for automatic MathML generation and a modern TeX distribution.
  • Documents built this way are fully WCAG compliant and offer strong screen reader support through MathML. You can safely ignore "missing alt text" warnings for these files, since they're legacy flags that don't recognize the accessibility data already embedded in the PDF.
  • Resources: Accessible LaTeX templates, ltx-talk documentation, a script to convert Beamer presentations to ltx-talk

If you're remediating a PDF with no source file

If math is saved as an image, screen readers can't access the underlying information, which happens often with scanned worksheets, handwritten notes, or whiteboard screenshots. Since these images carry no semantic information, assistive technology can't read or interpret the math.

• Use an AI tool to transcribe the equation. Upload a screenshot and ask it to "transcribe the expressions in the attached image using LaTeX."
• Add the transcription to the image's alt text or a nearby text description. This gives at least a minimally accessible version when the original source can't be updated.

Additional reference: Author guidelines for preparing accessible mathematics content (SIAM)

AI for complex graphs and figures: AI tools can help draft alt text for graphs and figures that are hard to summarize, giving you a starting point to review and refine rather than a final answer.

Converting scans of handwritten notes to accessible content

Handwritten notes shared as scans are image only files and are inaccessible as is. An AI tool can help convert them into an accessible format. Decide which workflow fits based on what's in the notes.

Scenario A: text heavy notes (mostly writing and equations, few or no diagrams)

1. Upload the scan (PDF or image) to your AI tool.
2. Ask it to transcribe the notes into clean HTML using only basic tags, no external CSS, inline styles, or classes, and to convert any math strictly to MathML with clear headings.
3. Copy the output into a new Canvas page using the page's HTML editor, then switch back to the standard view to confirm the text and math render correctly.

Scenario B: complex notes with diagrams

Part 1: Generate the content Ask the AI tool to provide the text of the notes as HTML (using MathML for math) with placeholders where diagrams belong, and to separately provide SVG code for each diagram. Keep the HTML text ready to paste into Canvas, and save each diagram's SVG code as its own file.

Part 2: Review and fix the diagrams AI generated diagrams can have small errors, like misspelled labels or rough curves. Open each SVG file in a vector graphics editor to fix labels, smooth out lines, and adjust colors for contrast. Export the cleaned version as SVG (best for quality) or PNG, checking "Responsive" if exporting SVG so it scales well on mobile.

Part 3: Publish to Canvas Upload your finished images to your course files. Create a new Canvas page and paste in the HTML text from Part 1, then switch to the visual editor, remove the image placeholders, and insert your finished images. Click each image to add descriptive alt text.

Alternative: a single all in one file If you don't need to edit the diagrams, ask the AI tool to output the notes as a single, self contained HTML file with embedded SVG diagrams, then upload that file directly to your course.