A practical introduction

AI is changing accessibility.

It can remove barriers in seconds. It can also make important decisions harder to understand, question or correct.

Understand AI accessibility ↓
01 / DEFINITION

What is AI accessibility?

AI accessibility concerns whether disabled and neurodivergent people can use AI, benefit from it and remain in control when it influences their lives.

The subject reaches beyond captions, alternative text and assistive technology. It also asks whether a person can understand an AI output, recognise uncertainty, challenge a recommendation, correct a mistake and recover safely.

That makes AI accessibility fundamentally a question of cognitive accessibility as well as technical access.

02 / THE MODEL

Three ways to think about it.

01

Accessibility of AI

Can people use the system?

Interfaces, conversations and multimodal experiences must work with different ways of seeing, hearing, moving, communicating and processing information.

02

Accessibility through AI

Can AI reduce a barrier?

AI may generate captions, explain difficult material, adapt information or support communication. The benefit depends on accuracy, context and meaningful human review.

03

Accessibility under AI

What happens when AI has influence?

When AI ranks, recommends, predicts or decides, people need understandable information, fair treatment and a realistic way to question the outcome.

03 / KEY QUESTIONS

What should a person be able to do?

Access

Perceive, reach and operate the system in a way that works for them.

Understand

Follow the interaction and make sense of the language, output and next step.

Evaluate

Recognise evidence, limitations, uncertainty and the possibility of error.

Stay in control

Correct, challenge, opt out or reach a person—and recover when something goes wrong.

Where it matters

The stakes rise when AI influences opportunity, support or essential services.

  • Recruitment and employment
  • Education and assessment
  • Banking, credit and insurance
  • Healthcare and care
  • Public and customer services
  • Workplace copilots and autonomous agents

Go deeper

Good intentions are not evidence of an accessible AI system.

Responsible teams set requirements early, include disabled people, test real journeys and monitor what happens after deployment.

Authoritative starting points