Can ChatGPT Audit Your Website for Accessibility?
Peter Bardenhagen
Technology Leader | Delivering Complex Digital Projects with Measurable Results
Background
The new ChatGPT models emerging like o1-Preview are the first to include reasoning capabilities. I wondered if it could audit a website for WCAG AA accessibility.
Steps to complete a WCAG Audit on a single web page
Poe (poe.com) offers a comprehensive platform that integrates multiple AI models, including ChatGPT and various text, image, and video analysis capabilities. While ChatGPT can also perform these audits independently, I've found success using the o1-preview model on either platform is the first version that has this reasoning capability.
The Process is Straightforward:
Within seconds, you'll receive an analysis based on the latest WCAG 2.2 guidelines. The report typically provides insights into:
Certainly! Based on the content provided for LinkedIn, I will perform an accessibility audit in accordance with the Web Content Accessibility Guidelines (WCAG) 2.2 Level AA. This audit will identify potential accessibility issues and provide recommendations to enhance the website's accessibility for all users, including those with disabilities.
Overview of WCAG 2.2 Level AA
WCAG 2.2 is organized under four main principles:
Audit Findings and Recommendations
1. Perceivable
1.1 Text Alternatives (1.1.1)
Issue: The content references images and multimedia (e.g., "We’re unlocking community knowledge in a new way" with AI assistance), but there's no indication of alternative text for images.
Recommendation:
1.2 Time-Based Media (1.2.x)
Issue: References to "live videos," "stories," and "newsletters" suggest multimedia content.
Recommendation:
1.3 Adaptable (1.3.1)
Issue: The structure of the content (headings, lists, forms) must be properly marked up to be interpreted by assistive technologies.
Recommendation:
1.4 Distinguishable
Issue: Text and interactive elements must meet minimum contrast ratios; content should not rely solely on color to convey information.
Recommendation:
2. Operable
2.1 Keyboard Accessible
Issue: All functionality must be accessible via keyboard alone.
Recommendation:
2.2 Enough Time
Issue: Users may need more time to read or interact with content.
Recommendation:
2.4 Navigable
Issue: Users need mechanisms to navigate, find content, and determine where they are.
Recommendation:
2.5 Input Modalities
Issue: Functionality requiring complex gestures may not be accessible to all users.
Recommendation:
3. Understandable
3.1 Readable
Issue: The site must declare the language used and manage any language changes.
Recommendation:
3.2 Predictable
Issue: Navigation and functionality should be consistent to avoid confusion.
Recommendation:
3.3 Input Assistance
Issue: Users should receive help in avoiding and correcting mistakes.
Recommendation:
4. Robust
4.1 Compatibility
Issue: Content and functionality must be compatible with assistive technologies.
Recommendation:
Additional Observations
Summary of Recommendations
Conclusion
Improving accessibility is crucial for providing an inclusive experience for all users, including those with disabilities. By addressing the issues identified in this audit and implementing the recommendations, LinkedIn can achieve greater compliance with WCAG 2.2 Level AA standards.
Next Steps
Note: This audit is based solely on the content provided and may not cover all aspects of the actual website's accessibility. For a comprehensive evaluation, consider conducting a full audit that includes testing with the live website and assistive technologies.
Final thoughts
Is it perfect? Not really. It offers a gauge of accessibility.
While exploring AI's capabilities in accessibility testing, I discovered both its potential and limitations. When asked to quantify a page's compliance as a percentage, the AI acknowledged its constraints and instead provided a structured analysis of strengths, weaknesses, and actionable next steps.
I was particularly impressed when discussing accessibility overlays. Rather than endorsing quick-fix solutions, the AI emphasized the importance of developing comprehensive remediation plans that integrate accessibility throughout the development process.
However, AI won't replace human expertise in accessibility auditing anytime soon. Its true value lies in augmenting human efforts by:
While automated tools can accelerate testing, thorough manual audits remain the gold standard for ensuring genuine accessibility. AI serves best as a complementary tool, enhancing—rather than replacing—human expertise in creating truly inclusive digital experiences.
Remember: True digital inclusion isn't about ticking boxes—it's about creating experiences that work for everyone. Whether you're just starting or enhancing existing practices, every step towards accessibility is a step worth taking.
Essential WCAG Resources
? W3C Web Content Accessibility Guidelines (WCAG) 2.2 https://www.w3.org/TR/WCAG22/
? W3C WCAG 2.2 Quick Reference https://www.w3.org/WAI/WCAG22/quickref/
? Understanding WCAG 2.2 https://www.w3.org/WAI/WCAG22/Understanding/
What's your experience with AI-powered accessibility testing? Let's continue the conversation in the comments below.
Managing Director at EstimatedStocks.com | 20+ Years in Tech & Finance | Empowering Investors with Data-Driven Insights
3 个月Thank you for sharing this insightful guide! The outlined approach provides a practical starting point for integrating AI into accessibility audits. I particularly appreciate the emphasis on combining AI-driven insights with manual testing and user feedback to ensure a comprehensive assessment. The suggestion to focus on one page or feature at a time for detailed analysis is a game-changer for efficient testing. Looking forward to exploring more about leveraging AI for accessibility improvements—especially in areas like dynamic content or advanced keyboard navigation. Great work! ??
Solution Architect | Passionate innovator focused on meaningful tech
4 个月Excellent tip: "cross reference against actual...". Something all Gen AI apps should do.