After a long winter, spring reminds us that brighter days are ahead. At AFB, your support helps ensure that people who are blind or low vision have the tools, research, and opportunities they need to thrive. Your gift helps build a more accessible world.
Give Today
Purple flowers in bloom with a sunset on the horizon behind them.

Introduction

An older woman works in an office on her laptop. She holds a phone in her other hand.

Artificial intelligence (AI) is becoming ubiquitous in the lives of many tech-savvy people. A December 2025 YouGov survey showed that 70% of U.S. adults had used generative AI at least once in the past year, with half of those adults using it at least weekly. Despite this widespread use, however, most Americans in this YouGov survey expressed low trust in AI; 41% stated they did not trust AI much or at all to provide information or make recommendations, and 53% did not trust AI much or at all to take actions or make decisions (YouGov, 2025).

Like their nondisabled counterparts, some people with disabilities are using AI to make daily tasks easier. New AI applications may also help people with disabilities access information; for example, AI-powered visual descriptions and captions can convert information from inaccessible to accessible modalities, helping overcome information access barriers. At the same time, though, Americans with disabilities likely share many of the reservations that nondisabled Americans feel about AI’s trustworthiness and safety.

In 2025, the American Foundation for the Blind (AFB) published a white paper detailing the consensus predictions of 32 experts on AI and disability. While the expert panel generally believed AI offers benefits for disabled people, they also expressed a variety of concerns. For example, the panel strongly agreed that automated job applicant screening and hiring systems could unwittingly screen out applicants with disabilities and that AI-driven health insurance decisions could disproportionately harm disabled people whose healthcare needs are not typical. The experts predicted that AI which describes pictures, in particular, could make mistakes that are not obvious to its users. There were also notable points of disagreement amongst the experts; for example, there was deep disagreement regarding the benefits and costs of autonomous vehicle (AV) development for people with disabilities, as well as AI’s potential to boost the productivity and integration of disabled workers (Silverman et al., 2025).

In the present investigation, AFB researchers surveyed a nationwide U.S. sample of adults with various disabilities who used AI at least occasionally, along with a comparison sample of American adults without disabilities. The survey was developed in collaboration with more than 15 disability advocacy organizations, and it was offered in plain language. Participants could choose to take the survey online, via phone, or via American Sign Language. The following research questions guided this study:

  1. How do people with and without disabilities differ in their use of mainstream AI tools, including use cases, frequency of use, and perceived helpfulness of the tools?
  2. Do disabled people experience access barriers when learning to use AI?
  3. How are job seekers with disabilities impacted by automated job screening assessments?
  4. How do disabled and nondisabled people feel about the privacy impacts of using AI?
  5. What suggestions do disabled and nondisabled AI users have for AI developers to improve AI systems?

In addition to these central research questions, the survey also examined experiences of those who had used AVs, those who used AI for psychotherapy or mental health support, and those who had experienced a healthcare denial. These questions were designed to address the fact that AI is playing a growing role in making healthcare decisions, including those surrounding treatment selection and insurance coverage (Hurley et al., 2024; Mello & Rose, 2024; Silverman et al., 2025).

In the survey, the term “AI” was defined broadly as “technology that can think, learn, and make decisions on its own.” This definition covered older automated systems for voice command recognition and word prediction, called “voice-activated AI” in the survey (e.g., Siri, Alexa) and automated captioning, as well as generative AI tools, automated tools that can read text or describe pictures (“visual description” tools), AVs, and automated job screening tools.