Accessibility isn’t just about disability, it’s about who gets included
I recently read a UN article about how AI is reinforcing gender inequalities.
Many of the underlying challenges are similar to web accessibility: they all come back to who is considered during design, development and testing.
Here are a few connections that stood out to me:
Biased data leads to biased outcomes
AI systems learn from existing data. If that data reflects historical inequalities, the system can reproduce them. The well-known case of Amazon’s recruitment tool is one example - it learned to favour male candidates because it was trained on historical hiring data. From an accessibility perspective, this is a reminder that designing or testing with only a narrow group of users also produces exclusion.
Representation matters during development
The UN article highlights that women remain underrepresented in AI development. Accessibility has long recognised that involving people with different lived experiences throughout design and testing leads to better products. Diverse teams alone are not enough, but diversity in research participants, testers and decision-makers helps uncover barriers that might otherwise go unnoticed.
Participation depends on feeling safe
The article also discusses how AI-generated abuse and online harassment disproportionately affect women. Accessibility aims to enable participation in digital spaces. While it doesn’t directly address online abuse, it shares a similar goal: ensuring people are able to access and participate in digital services without unnecessary barriers. Safety is one factor that influences whether people can truly participate.
Inclusion goes beyond compliance
Meeting accessibility standards is essential, but it doesn’t automatically mean a digital product is inclusive. As AI becomes part of more websites and applications, accessibility professionals have an opportunity to contribute beyond WCAG compliance by asking broader questions:
Who was represented in the data?
Who participated in user research and testing?
Which users might experience unintended barriers?
Are we designing for real diversity rather than an “average” user?
For me, the article reinforced something I’ve learned through accessibility work: inclusion isn’t achieved by assuming who the user is. It starts by recognising that people experience technology differently and designing with that in mind.
What other connections do you see between responsible AI and accessibility?