AI Model Wiki

    Skip to main content
    14 min read
    Lesson 2 of 5

    AI-Powered Alt Text & Image Descriptions

    From the course AI for Accessibility & Inclusive Design

    Alt text is the most fundamental accessibility feature for visual content. AI vision models can now generate contextually appropriate descriptions at scale.

    How AI Alt Text Works

    Vision-language models (GPT-4o, Gemini, Claude) analyze images and generate natural language descriptions. The quality has improved dramatically:

    • 2020: "A dog on grass" (generic, minimal context)
    • 2024: "A golden retriever puppy playing fetch with a red ball in a suburban backyard on a sunny afternoon" (rich, contextual)

    Writing Effective Alt Text Prompts

    The key is context. The same image needs different alt text depending on where it appears:

    For a pet adoption site:
    "Describe this image for a screen reader user on a pet adoption website. 
    Focus on the animal's breed, age estimate, temperament cues, and physical features."
    
    For a photography tutorial:
    "Describe this image for a screen reader user on a photography education site.
    Focus on composition, lighting technique, camera angle, and artistic choices."
    

    Implementation Approaches

    Batch processing: Generate alt text for existing image libraries

    • Process images through a vision API
    • Store descriptions in your CMS alongside images
    • Human review for quality and accuracy

    Real-time generation: Generate descriptions on demand

    • User uploads or encounters an image without alt text
    • API call generates description in <2 seconds
    • Cache results to avoid redundant API calls

    Browser extensions: Tools like Be My Eyes and Microsoft's Seeing AI provide on-device image descriptions for users browsing any website.

    Quality Guidelines

    Good AI alt text should:

    • Be concise (125 characters for simple images, up to 250 for complex ones)
    • Describe function, not just appearance ("Submit button" not "blue rectangle")
    • Avoid "image of" or "picture of" — screen readers already announce it's an image
    • Include text visible in the image
    • Describe charts and graphs with data trends, not just "a bar chart"

    Handling Complex Visuals

    For charts, infographics, and diagrams, use a two-tier approach:

    1. Short alt text: Brief summary ("Sales grew 40% year-over-year from 2022 to 2024")
    2. Long description: Detailed data table accessible via aria-describedby or a linked page

    Key Takeaways

    • •Context-aware prompting produces dramatically better alt text than generic descriptions
    • •Use batch processing for existing libraries and real-time generation for new uploads
    • •Good alt text is concise, functional, and avoids redundant phrases like 'image of'
    • •Complex visuals need both short alt text and detailed long descriptions with data
    0 of 5 lessons marked complete

    Progress is kept in this browser only.