This article responds to an article by Equalify’s founder, “A11y’s AI Future (The Future is Now)“. Equalify encourages open conversation and hopes to publish all viewpoints.
The recent article by Blake Bertuccelli-Booth, “A11y’s AI Future (The Future is Now)”, suggests that AI-driven tools, particularly Apple’s new AI tool, could significantly shift how accessibility is approached, potentially reducing the reliance on traditional tools like screen readers. While the excitement for AI’s potential is palpable, the article overlooks several critical factors that make web accessibility a complex and nuanced field.
AI Cannot Replace Semantic Structure in Accessibility
The article imagines a future where generative AI might render traditional screen readers obsolete by summarizing content and guiding users through tasks via natural language processing. However, this perspective underestimates the fundamental importance of semantic code in accessibility.
Semantic HTML provides structure and meaning to web content, which screen readers and other assistive technologies rely on to convey information effectively. Tags like <h1> are not merely decorative; they signal the beginning of a section and help users navigate content hierarchically. AI, no matter how advanced, cannot replace these essential web standards. Instead, AI should be seen as a tool to enhance the user experience by working in tandem with the semantic code, not by substituting the very foundation that ensures accessibility for all users.
Overreliance on AI Could Undermine Accessibility Standards
The article’s suggestion that AI tools can interpret context without relying on semantic tags could inadvertently encourage web developers to deprioritize proper coding practices. Getting developers to do this now is often an uphill battle in this industry. This could lead to poorly structured websites that exclude users who depend on assistive technology, like screen readers and speech-to-text tools.
Screen readers have a long history, rooted in decades of community input, research, and development. The first screen reader was developed by IBM in 1986, and since then, these tools have evolved significantly to become indispensable for millions of users worldwide. Similarly, the Web Content Accessibility Guidelines (WCAG) were first introduced in 1999, setting the foundational standards for creating accessible web content. These guidelines, alongside screen readers, are the result of years of collaboration and refinement, ensuring that digital spaces are navigable for all users.
It’s important to recognize that AI tools are not infallible. They are prone to biases and errors, especially when interpreting complex language or visual content. A page summarized by AI might miss critical context or misinterpret the content, leading to misinformation or an inaccessible experience for users.
Moreover, while large language models (LLMs) like GPT are powerful, they are far from capable of replacing mission-critical tools like screen readers. The significant limitations and challenges these AI systems face are unlikely to be resolved shortly, if ever. This raises a deeper concern about the current wave of AI enthusiasm. The real danger is that we might start relying on these half-baked, cost-effective chatbots to replace systems that, while labor-intensive, are fully functional and essential.
This trend is particularly worrisome because it presents complex, often nuanced problems as solvable by simplistic, sometimes one-line code solutions. These solutions are frequently touted to non-technical leaders as quick fixes for accessibility issues. However, simply implementing these on a website does not address the underlying problems, and in many cases, it can interfere with a user’s assistive technology, introduce additional accessibility barriers, and unwittingly inject security vulnerabilities. Accessibility requires careful consideration and human expertise, not just the application of superficial, bolt-on solutions after the fact.
Accessibility is fundamentally about providing access for humans, and that requires human intervention. AI is not the panacea it is often perceived to be, and as a screen reader user for over two decades, I find this notion deeply troubling. The idea that AI can replace the nuanced understanding and intentionality that human oversight brings to accessibility is not only misguided but also scary and problematic.
The Role of Accessibility Professionals: Beyond Fact-Checking
The article asserts that accessibility professionals will primarily become fact-checkers of AI outputs. While identifying AI errors is vital, this view reduces the scope of our role far too much. Accessibility professionals are not just fact-checkers; we are designers, testers, educators, and advocates for inclusive web practices. Our work ensures that technology serves everyone, regardless of their abilities.
Our roles extend far beyond verifying AI-generated content. We conduct workshops, providing hands-on training that gets into the nitty-gritty details of accessibility practices. We consult on very technical problems, helping developers implement complex solutions that require deep expertise in both accessibility standards and the specific technologies at play. These are not tasks that can be automated or replaced by AI; they require human judgment, experience, and a nuanced understanding of the diverse needs of users with disabilities.
By educating developers, auditing websites, and pushing for adherence to accessibility standards, we ensure that accessibility is not just an afterthought but a core consideration in the design and development process. Our involvement is crucial in creating digital experiences that not only follow regulations and conform to internationally recognized standards but also genuinely usable by all people.
AI Should Enhance, Not Replace, Accessibility Tools
AI has the potential to significantly improve accessibility by providing new ways for people with disabilities to interact with digital content. However, it’s crucial to position AI as a supplement to existing tools, not a replacement. Screen readers and other assistive technologies are built on years of development and community feedback, and they remain indispensable for many users.
As accessibility professional and developer Alex Stine insightfully noted, “AI is a buzz term right now and accessibility pros are often not early adopters, but for good reason. I also don’t want to discount what AI can do and has already done for people with disabilities; you just have to walk that line of enhance vs. end all be all.” Stine’s words capture the delicate balance we must maintain—AI should be seen as a tool that enhances accessibility, not as a one-size-fits-all solution that renders established tools and practices obsolete.
The notion that screen readers might be replaced by AI overlooks the diverse ways users interact with technology. For many, the predictability and consistency of traditional assistive tools are essential for navigating digital spaces efficiently. AI’s role should be to enhance these experiences, not to undermine or replace them.
A Balanced Approach to AI in Accessibility
The excitement around AI is understandable, but it’s vital to approach it with a balanced perspective. AI can bring incredible benefits to the field of accessibility, but it must be implemented thoughtfully, with a strong foundation in established accessibility principles. Semantic coding, for instance, is not just a fallback—it is the bedrock upon which accessible digital experiences are built.
The future of accessibility will indeed involve AI, but it will also require the continued expertise of accessibility professionals who understand the intricacies of both technology and user needs. Our role is not diminishing; it’s evolving to ensure that new technologies are implemented in ways that truly benefit all users.