AI for Accessibility?

Image descriptions (alt text) help make visual content accessible, not only for blind students but also for those experiencing technical difficulties. As subject matter experts, faculty are best positioned to describe the visual content in their courses. You know why you selected a specific image and what it conveys to your students.

The Role of AI in Drafting Alt Text

Brevity and clarity are essential for image descriptions, though complex visuals such as charts and graphs can be challenging to explain. Artificial intelligence (AI) tools can generate draft descriptions that you can then refine and adapt for your course. As with any other AI-generated content, review and adjust the image description for accuracy and to align with your course.

Let’s see how a few tools compare using the image in this article.

Microsoft Word, for example, incorporates AI. Under the Picture Format tab, selecting the Alt Text option prompts Word to generate a suggested description. However, results are often limited. For the image used here, Word produced only: “A graph of multiple colored vertical lines.”

Arizona State University (ASU) offers a growing collection of tools with their ASU Online AI Labs. Their Image Accessibility Creator, one of our favorites, produces multiple outputs including long descriptions, alt text, transcribed text, and a legend. 

For our sample image, combining the short alt text with the transcribed data produced this description:

Bar chart displaying average scores for different issue types in percentages.

• Color : 83% 
• Embedded Media : 100% 
• Images : 79% 
• Links : 77% 
• Page Headings : 84% 
• Page Structure : 91% 
• PDF : 70% 
• PowerPoint : 92% 
• Table Headers : 68% 
• Video Captions : 59% 
• Word : 67%

AI Tools for Image Descriptions

CSUN provides access to Microsoft Copilot and ChatGPT, both of which can generate image descriptions. However, the initial outputs were overly detailed. Using the refined prompt “Please provide a concise description of this image for a blind person” resulted in far more effective descriptions.

ChatGPT Results

This image is a bar chart comparing accessibility issue types by average score (0–100%). Each colored bar represents one issue type.

• Highest score: Embedded Media (100%).
• Other strong areas: PowerPoint (92%), Page Structure (91%), Page Headings (84%), Color (83%).
• Middle range: Images (79%), Links (77%), PDF (70%), Word (67%), Table Headers (68%).
• Lowest score: Video Captions (59%).

In summary, most areas score between 70–90%, with video captions as the weakest point and embedded media as the strongest.

Microsoft Copilot Results

This image is a bar chart titled “Issue Type Current”, showing average accessibility scores (in percentages) for different types of digital content issues. Each issue type is represented by a colored bar.

Here’s a concise summary of the key data:

• Highest score: Embedded Media – 100%
• Lowest score: Video Captions – 59%
• Other notable scores:
  • Color: 83%
  • Images: 79%
  • Links: 77%
  • Page Structure: 91%
  • PowerPoint: 92%
  • PDF: 70%
  • Word: 67%
  • Table Headers: 68%

By thoughtfully combining your expertise with AI-generated drafts, you can create image descriptions that enhance accessibility and ensure all students can fully engage with your course materials.