The AI Quagmire
Benefits, Risks, and Aspirations Through a Disability Lens
Artificial intelligence (AI) is influencing all aspects of our lives, shaping how people learn, work, communicate, and even get around. For people with disabilities, AI promises the moon: making information accessible, supporting independence, and reducing longstanding barriers. At the same time, AI can be inaccessible, it can make unfair decisions, it can make mistakes, or it can put people’s privacy at risk.
The American Foundation for the Blind (AFB) published findings from a survey comparing how people with and without disabilities are using AI. The survey also looked at the benefits and challenges people with disabilities have when using AI or automated systems. AI developers, employers, educators, policymakers, and users themselves can use the findings to help make sure AI includes everyone.
About AFB’s research on this.
AFB surveyed 1,735 adults in the United States between July 1 and October 15, 2025. There were 1,070 people with disabilities and 665 people without disabilities who took the survey. The respondents could take the survey online, over a video call, or in American Sign Language.
Researchers asked participants about how they learned to use AI, what kinds of AI they use, how AI affects work and education, and what concerns they have about privacy, healthcare, hiring systems, and autonomous vehicles (AVs).
What did the researchers learn?
- Most participants said they learned to use AI on their own.
- Overall, people with and without disabilities used AI for many of the same things, like helping with research and writing. Many users with disabilities also used AI to turn print, pictures, or speech into digital text they can easily read.
- Most AI users said AI makes their lives easier.
- However, some people said that AI trainings were not accessible to them, while others said that they had trouble completing automated job assessments (like a computerized typing test or interview) with their assistive technology.
- Some people who used AI for describing pictures or captioning speech said the AI made mistakes and gave them the wrong information, like misreading a medication label. Others said AI voice assistants misunderstood them or failed after updates.
- In the survey, people with disabilities couldn’t get a medication or procedure almost three times as often as people without disabilities. This matters because healthcare decisions are often made by AI.
- Regarding AVs, disabled people who cannot drive thought they were more important to build than people who drive and people without disabilities. Everyone agreed that public transit is still very important.
- Blind and low-vision (BLV) people were especially excited about AVs, but about half of the BLV people who rode in an AV said the ride wasn’t fully accessible, while people who use wheelchairs or walkers usually still cannot use AVs.
- Most respondents said they would still prefer getting help from a human instead of AI to handle private information, and they were much more comfortable using AI to read something private if the AI didn’t save any information or send it to a tech company. Respondents generally thought AI was less private than humans and that privacy was important.
Why is this a problem?
When AI systems aren’t accessible or when they make biased decisions, disabled people are left out of important things, like jobs, healthcare, and transportation. Disabled people can also be harmed when AI gives them the wrong information. If an AI system saves or shares information, disabled users may have to choose between using AI for access and protecting their privacy.
Who needs to take action to change this?
Policymakers at the national, state, and local levels, AI developers, technology companies, employers, educators, healthcare systems, insurers, transportation providers, researchers, and advocates.
What changes does AFB recommend based on the research?
- Ensure that all platforms that integrate AI are fully accessible to and usable by people with disabilities.
- Improve privacy and data security practices to increase trust in AI products and support the use of AI when sensitive information is involved.
- Improve the accuracy of AI outputs and tell users how accurate those outputs are.
- Ensure that AI used in high-impact areas, like employment and healthcare, is well-trained and monitored so it does not make inappropriate decisions that affect people with disabilities and other groups.
- Establish governmental guardrails and policies that promote fairness in high-impact use cases, require data privacy and security, and ensure accessibility for people with disabilities.
Download the Report Read the Entire Report Online
Glossary of Terms
Accessible: Designed so people with disabilities can use it. In this report, that includes things like readable training materials, keyboard access, compatibility with assistive technology, captions, and audio information.
Artificial intelligence (AI): In this study, AI means technology that can think, learn, and make decisions on its own. The report uses this term broadly to include voice assistants, chatbots, captioning tools, visual description tools, autonomous vehicles, and automated job screening systems.
Autonomous vehicle (AV): A vehicle that can drive itself using automated systems. The report examined both the potential benefits of AVs and the barriers riders faced.
Automated job screening: A job application step where the person applying interacts with a computer instead of a human.
Blind and low-vision (BLV): A term used in the report for blind and low-vision participants. The report notes that identity-first language was intentionally used for this group.
Captions: Written text that shows spoken words in text form. The report found that captions can improve access, but caption errors can also cause harm.
Human oversight: A person reviewing, guiding, or double-checking what AI does. The report’s recommendations show that this is especially important in high-impact settings.