Artificial intelligence (A.I.) has the potential to dramatically improve web accessibility — but as with any new technology, thoughtful adoption is key.
Generative A.I. refers to tools that are capable of creating new content. Currently, ChatGPT and DALL-E are the best-known examples, but hundreds of companies are scrambling to introduce new generative A.I. tools.
At the Bureau of Internet Accessibility, we’ve used artificial intelligence as part of our automated web accessibility scans for years. AudioEye uses advanced algorithms to automatically fix common accessibility failures while providing guidance for issues that require human judgment to fix.
As generative A.I. improves, it’s possible that the list of issues that require “human judgment" will shrink — and that could have a tremendous impact for both website owners and internet users with disabilities.
Most digital accessibility issues can be addressed easily with clean code and thoughtful content creation. However, many “easy" fixes still take time to implement, particularly when humans need to be involved.
The Web Content Accessibility Guidelines (WCAG) function as the international standards of digital accessibility. WCAG includes a number of requirements that require a subjective approach, which can create busywork for developers, designers, and writers.
For example:
Generative A.I. may be able to address these challenges. OpenAI’s GPT-4 can accurately describe many images, which could reduce the need for manual alt text. Speech-to-text tools may be able to recognize a greater variety of human voices, which could eventually allow for accurate realtime captions and transcripts.
But relying solely on A.I. isn’t ideal. An A.I. tool designed to describe an image might miss important information about the image’s function on the page — and while automatic captions can be helpful, current tools are limited. For example, YouTube’s automatic captions are only about 60-70% accurate.
Related: Is A.I.-Generated Content Bad for Accessibility?
Currently, A.I. can address many WCAG requirements that have simple “pass or fail" rulesets. That includes issues like keyboard traps, low-contrast text, and improper use of semantic HTML. Over time, we expect that to change, and more WCAG criteria will be fixable with A.I.
But for optimal results, A.I. tools need to be designed specifically for accessibility — and every accessibility fix should be manually tested to ensure that it actually improves on-page experiences for users.
At the Bureau of Internet Accessibility, we recommend combining automated tools with manual review to improve WCAG conformance (and compliance with the Americans with Disabilities Act and other non-discrimination laws).
That approach is recommended by the Department of Justice (DOJ), the World Wide Web Consortium (W3C), and other institutions that monitor digital accessibility.
Related: How to Check WCAG Compliance: A Quick Guide
Currently, there’s no substitute for an accessibility-first mindset. When you build products for all users — regardless of their abilities or conditions — you enjoy the full benefits of inclusive design: a larger audience, a stronger brand, and lower long-term costs for development.
To learn about the basics of accessibility, download our free eBook: Developing the Accessibility Mindset.
And if you’re ready to build an accessibility strategy for your business, we’re here to help. Get started with a free automated website analysis or send us a message to connect with an expert.