Accessibility includes functional accessibility (can a person with a visual impairment use this tool?) and structural accessibility (can someone in a rural, low-bandwidth region afford or reach this service?). It also requires that automation not systematically strip opportunities, such as employment or essential services, from already marginalised people. 

An automated system (such as a self-service kiosk at the airport or a smart thermostat) is accessible when its physical and digital interfaces are usable by people with physical disabilities or who suffer from reduced cognitive capacity, lower economic status, or who are marginalised by language, geography, or age. 

For these kinds of systems, buttons should be at a height reachable for someone in a wheelchair, audio instructions provided for people who cannot see the screen, and steps clear enough for people with a cognitive disability to follow.

By contrast, the accessibility of an AI-powered system is not just about the "buttons" on the screen; it’s about how the system thinks and communicates. 

For example, does the system recognise speech from someone with a stutter or an accent? Does it recognise faces with different skin tones or assistive devices? Can it summarise a complex legal document into "Plain English" for someone with a learning disability? Is it programmed not to "filter out" or penalise people because of their disability during tasks like job screening.

Why it matters

In a world increasingly mediated by automation and AI - from job applications to healthcare - lack of accessibility creates a digital divide. Society thrives when every citizen has the agency to use essential tools.

Excluding a significant portion of the population (roughly 16% globally) from AI advancements stifles economic participation and reinforces social marginalisation. 

Historically, accessibility has been treated as an afterthought in technology development, and AI risks repeating and compounding this pattern at far greater scale and speed. 

Impacts & consequences

• Employment exclusion. AI systems trained on historical hiring data are likely to replicate and amplify existing patterns of exclusion. Due to high unemployment rates among people with disabilities, employment datasets used by AI tend to underrepresent them, making discriminatory outcomes more likely.

• Functional barriers. Many AI systems rely on voice commands for operation, which can be challenging for individuals with speech disabilities or accents the technology fails to recognise. AI-powered websites and apps may not be designed with the needs of individuals with visual or motor impairments in mind, creating new access barriers.

• Biased representation. AI image generation has been found to reproduce stereotyped depictions of disability - for instance, generating images of autistic people as predominantly white males - reflecting narrow and biased training data.

• Automated discrimination. AI-powered automated decision-making systems may unintentionally discriminate against people with disabilities by not accounting for their unique needs and circumstances.

• Legal and reputational risk. Organisations deploying inaccessible or discriminatory AI face growing legal liability under frameworks such as the Americans with Disabilities Act (ADA) and the EU AI Act.

Causes & contributing factors

The inaccessibility of AI and automation system may stem from one or more factors, including: 

• Training data is often drawn from the most digitally active populations, leaving underrepresented groups poorly served by the resulting models. 

• Design processes rarely include disabled or economically marginalised users, producing tools built around an assumed "default" user. 

• Commercial incentives drive development toward profitable markets, neglecting communities that cannot pay premium prices. 

• Infrastructure inequality (aka the "digital divide") means that even accessible tools remain out of reach where internet connectivity, hardware, or electricity is unreliable. 

• Automation-driven job displacement disproportionately hits lower-skilled workers, deepening existing socioeconomic divides.