Innovation for Access
AI and Its Use as Assistive Technology for People with Disabilities
Supplementary Research | August 2026
This report analyzes AI and Assistive Technology using data from the survey, The AI Quagmire: Benefits, Risks, and Aspirations Through a Disability Lens. Researchers from the American Foundation for the Blind surveyed U.S. adults with and without disabilities about AI's use and impact. Building on these findings, Innovation for Access explores AI as assistive technology (AT) for people with disabilities, focusing on AI-generated visual descriptions, automated captions, and Augmentative and Alternative Communication (AAC). AI is now common in classrooms, workplaces, and other environments, and it may be well-suited to support those with disabilities as AT integration continues.
Download the Innovation for Access report or access all contents below.
Download the Innovation for Access Report
Authors: William E. Hanuschock, Ed.D. | Arielle M. Silverman, Ph.D.
Contributors: Sarahelizabeth J. Baguhn, Ph.D. | Mana Hashimoto, M.A. | Carmel Heydarian, M.S. | Omar Khan, B.S. | Mei-Lian Vader, M.S. | Angie L. Whistler, M.S.Ed.
Table of Contents
Augmentative and Alternative Communication (AAC)
Recommendations for Innovation
Language Note
In this report, we have used both person-first language (“people with disabilities”) and identity-first language (“disabled people”) when describing disability. This is an intentional choice meant to honor differing preferences within the broader disability community, an approach supported in research literature (e.g., Dunn & Andrews, 2015). In addition, we have used identity-first terms when specifically describing three disabilities: “BLV people,” meaning blind and low-vision people; “D/HH people,” meaning deaf/Deaf and hard-of-hearing people; and “autistic people.” This choice reflects the strong preference for identity-first language voiced by many members of these particular communities.
Acknowledgments
We thank Sarah Malaier and Stephanie Enyart for developing the recommendations at the end of this report and for offering feedback on earlier drafts. We are also grateful to the working group of 19 disability organizations whose input shaped the survey questions and design. Finally, we thank the Ford Foundation for supporting the American Foundation for the Blind’s efforts on AI and workplace technology to ensure equal rights for people with disabilities.
Introduction
Artificial intelligence has become commonplace in a wide range of environments, such as classrooms, boardrooms, and workplaces, for many Americans. Because AI can readily convert information from one modality to another, AI may also be well-suited to act as assistive technology. Assistive technology (AT) is technology intended to “increase, maintain, or improve the functional capabilities of people with disabilities” (Assistive Technology Act, 2004). As innovations in AT continue to develop, the integration of AI will become commonplace to support those with disabilities.
There are many innovations embedding AI in AT. This report will focus on three common areas. First, new visual image description tools (also called visual description or image description in this report) utilize AI to generate verbal descriptions of pictures or videos and conversion of printed text into digital text form. This technology is often designed specifically for blind and low-vision (BLV) users, but people with other types of disabilities, such as difficulty with facial recognition or visual processing, may also benefit. Second, automation increasingly replaces human captioners when creating captions that convert speech to text. These captions are especially valuable for users who are deaf/Deaf or hard of hearing (D/HH), but people with many other types of disabilities may routinely use captions as assistive technology.
Furthermore, people without disabilities often utilize visual descriptions or captions to improve comprehension or to support multitasking, among other reasons. Finally, augmentative and alternative communication (AAC) applications, utilized by some people with speech or physical disabilities and some autistic people, have begun using AI to generate synthetic voices and to better predict what a user is typing, and AI’s role in AAC will continue to increase over time. Taken together, these technologies are utilized by people with a wide range of disabilities, including BLV and D/HH people, those with speech disabilities, and those with visual and auditory processing disabilities.
In the following report, we examine the use of AI-generated visual descriptions and captions, as well as AAC use, reported in a 2025 survey conducted by the American Foundation for the Blind (Silverman et al., 2026). The larger survey comprised 1,735 participants, including 1,070 with disabilities and 665 without disabilities. Complete demographics for the full sample are available in AFB’s March 2026 report titled, The AI Quagmire: Benefits, Risks, and Aspirations Through a Disability Lens (Silverman et al., 2026). The survey asked users of these AI-enabled assistive technologies to describe use cases, frequency of use, perceptions of the quality of the technology used, and suggestions for improvements. Additionally, visual description users reported on their attitudes toward using AI to read print containing sensitive or private information.
AI Use in Visual Description
Visual Description Users
There were 705 participants (511 disabled, 194 nondisabled) who answered “yes” or “I am not sure” to a question about whether they had used AI applications to “turn pictures into text; for example, to describe pictures or videos, or read text aloud.” Of these, 10 participants skipped later questions about AI visual description and were excluded. Another 73 participants were excluded from these analyses because they indicated on at least one follow-up question that they actually did not use AI visual descriptions. Thus, data were analyzed from 622 AI visual description users, 458 (74%) of whom were disabled. Of the 458 disabled users, 290 were BLV, while 168 were sighted with other disabilities. Most of the BLV participants in the full study (79%) used visual description.
Use Cases and Frequency
AI visual descriptions were used for several tasks. For BLV users, reading text was the most common use case (93%), followed by describing photos taken by oneself or others (78%), describing scenes (78%), and figuring out object details, such as what is in a can (76%). For sighted users, the most common use was to read text, though at a lower frequency (60%), followed by describing scenes (44%), describing photos (42%), and describing videos (40%). BLV participants used AI to describe videos at a similar rate to sighted participants (39%). Additional use cases for AI visual descriptions included reading signs and identifying the mood or emotions of people in photos.
Among BLV users, 51% reported using AI for visual descriptions at least once per day, while 36% of BLV participants used it weekly, and the remaining 15% used it less than once a week. In contrast, only 9% of sighted users used these descriptions daily, and 28% used them weekly, with most sighted participants (73%) using AI for visual descriptions less than once a week.
Visual Description Accuracy and Errors
Overall, 8% of visual description AI users thought the descriptions were extremely accurate, 57% thought they were mostly accurate, 29% thought they were somewhat accurate, 1% thought they were not at all accurate, and 4% stated they were not sure how accurate the descriptions were. Disabled users, including BLV users, thought that they were more accurate than users without disabilities. Disabled users were likelier than their nondisabled counterparts to rate the descriptions as “mostly accurate” (59% vs. 52%), while nondisabled visual description users expressed more hesitancy, rating the AI-generated descriptions as “somewhat accurate” more often (36%) than disabled users (27%). Disabled users were also slightly more likely to say the AI descriptions were “extremely accurate” (9%) compared to 7% of nondisabled users.
With regard to errors, 90 AI visual description users said the AI made mistakes that hurt them. This included 61 BLV users (21% of the BLV group), compared with 14 sighted disabled users (9% of the sighted disabled group) and 15 nondisabled users (9% of the nondisabled group). Thus, BLV users were much more likely than sighted users to be negatively affected by an AI visual description error. This may be due to not being able to visually confirm the accuracy and fidelity of the image, scene, or item in question. Additionally, as the quotes below illustrate, some BLV users use visual descriptions in higher-stakes situations, such as reading financial or medical information, and sighted users might not have those use cases as often.
These harms ranged from legally harmful to medically dangerous. One participant used an AI tool to assist with reading and completing tax forms, which were not screen reader accessible. They wrote that the AI “provided some very incorrect information that would have caused serious tax filing mistakes if I had not checked the information with a human.” Another participant recounted their experience using AI to read their printed medication label:
Using AI to read package instructions on a tube of topical medication and when I asked what the directions for use were specifically, AI told me to ‘chew 4 tablets 2 times per day.’ Luckily, this was an obvious error, but for now I will NOT be using AI for these tasks [and] will always confirm with a human.
Privacy Considerations
A common assumption, based on a historical lack of access, is that people with disabilities are willing to give up privacy in order to gain access to information and their environment. To assess actual perceptions about the user privacy preferences, all visual description users were asked about using AI to read sensitive information in two scenarios. First, these participants indicated their willingness to read sensitive information with an AI tool that did not save or share the images. In this scenario, regardless of disability status, 60% of users stated they would prefer to use AI over a human reader. However, when the AI tool would share images with a tech company, only 17% of disabled and 28% of nondisabled users said they would prefer the AI. Thus, participants would consider using AI for private information if the AI kept the information “on device” or deleted it immediately. Participants, especially those with disabilities, were concerned about the privacy of sensitive information, and many reported reluctance to use AI tools for reading or processing this information if the tool were likely to share their information with others. This result tracks a finding from the full study in which a significant majority of disabled participants reported that privacy is equally important as or more important than the efficiency and independence that AI provides (Silverman et al., 2026).
Ideas for Further Innovation in Visual Description AI
Participants shared how they thought AI visual descriptions could be made better. Four themes were identified: accessibility improvements, feature suggestions and improvements, increased accuracy of the software’s responses, and privacy concerns. The largest theme among responses was accessibility; participants wanted AI tools to have a good nonvisual interface for taking pictures. One participant shared:
I would like instructions on how to make the image clearer. For example, move the camera left or right. Also, I’m always being told there isn’t enough light when I’m sure there should be. Turning on the flashlight of my phone automatically would be helpful since I am totally blind.
Another participant indicated, “It would be nice if you could just point your phone in a direction and the scene would automatically be described to you. The same goes for text.”
The next theme was new features and improvements. One participant wrote, “Make tutorial videos to help people learn to use it. We didn’t grow up with this AI technology. I’m sure there are features I’m not aware of. For example, I’d like to learn about AI used for creating pictures.” Another added, “I wish it had a larger database for reading barcodes on packages. Sometimes it won’t read the packages from a discount store like Save A Lot or Aldi.” Other features that participants wanted their AI software to have include more comprehensive descriptions, support for image editing, more control over the level of detail in the descriptions, and environment navigation.
A third theme was increased accuracy. One person shared, “I am not really sure. Just more accuracy would be very beneficial. I’m less worried about data [privacy] because I think all of our data is out there anyway, so the hallucinations are my main gripe with AI.” Although most users perceive the visual description outputs as being mostly accurate, there is clearly room for improving the accuracy of outputs, especially for those users who cannot independently verify what the visual description tool produces.
Finally, some participants discussed privacy concerns here. One participant indicated, “Prioritize the privacy element,” and another said, “Privacy is imperative. It feels like they are snooping…” Another added, “I believe, as a conservative, that artificial AI sort of affects privacy. What if you’re trying to view a bank statement or view private information, you don’t want the AI gleaning that.” These comments echoed the finding that most users would prefer not to have sensitive information shared with AI companies. The remaining BLV participants gave comments that did not fit into these themes, such as stating they had no suggestions.
Among sighted participants, several gave feature suggestions. A sighted participant with a physical disability suggested, “Learn from questions I ask about the level and type of detail I want. Parental controls are fine, but I should also have the ability to have virtually any photo described without AI telling me there is adult content and thus it can’t do it.” A nondisabled participant suggested ways to improve a visual description chatbot: “Continue to enrich the user experience with more description, more interaction, more ‘human’ sounding voices.” Increased accuracy also came up in the sighted group, with one participant indicating simply, “It should be more accurate!” Privacy concerns were raised at a similar frequency to BLV participants, and some sighted participants discussed accessibility concerns, but not as frequently as BLV participants.
When comparing both groups, those in the BLV group wanted an increase in accessibility features at a greater rate than sighted people. On the other hand, sighted participants made more feature suggestions. In relation to the features suggested, BLV participants wanted more position-related feedback (e.g., directions on where to move the phone to assist with taking a picture), whereas sighted participants wanted more options related to output (e.g., emotion, storytelling, etc.). Both groups expressed interest in improving the accuracy of the visual descriptions.
AI Use in Captions
AI Caption Users
There were 933 participants (602 disabled, 331 nondisabled) who answered “yes” or “unsure” to a question about whether or not they had used AI captions to “show spoken words in text form.” Of these, 20 participants did not answer more specific caption use questions, and they were excluded. Another 71 participants were excluded because they indicated in at least one later response that they did not actually use AI captions. The final sample thus comprised 842 caption users (527 disabled, 315 nondisabled). The disabled group included 54 D/HH users (61% of the full D/HH sample), and 473 fully hearing users with other disabilities.
Use Cases and Frequency
For both hearing and D/HH participants, it was common to use AI captions on recorded videos (85% of D/HH and 70% of hearing participants) and captions on video meetings (70% of D/HH and 48% of hearing). Hearing participants were likelier to download AI transcripts after meetings (40%) than D/HH participants (30%). Other common uses, especially for D/HH participants, were captioning face-to-face conversations or lectures and captioning live events.
Overall, the D/HH participants used AI captions more frequently than the hearing participants. 54% of the D/HH participants used them at least once a day, 24% used them weekly, and 22% used them less than once a week. Hearing participants were much less likely to use AI captions daily (20%) and more likely to use them weekly (37%) or less than once a week (42%).
AI Caption Setup
Sometimes AI captions are available automatically, whereas in other instances the user must request or set them up. Users may experience time or effort burdens when they have to set up or request captions. Caption users were asked how often they have to set up or request the AI captions they use. Among D/HH users who use captions daily, 47% reported that they set up or request captions at least half the time they use them. This number was much higher than it was for D/HH users who use captions less often; among those who use captions weekly or less, only 22% reported setting up or requesting captions at least half the time, and among hearing users, 27% reported setting up or requesting captions at least half the time. D/HH users who rely on captions for daily access may experience greater burden from having to set up or ask for captions frequently, compared to hearing users and infrequent caption users. Since the survey did not ask D/HH users for more details about the impact of setting up or requesting needed captions, this finding should be interpreted with caution.
AI Caption Accuracy and Comparison to Human Captions
When asked about their perception of AI caption accuracy, only 4% of participants thought AI captions were extremely accurate, 53% rated them as mostly accurate, 42% rated them as somewhat accurate, and 1% rated them as not at all accurate. In contrast to the AI visual description findings, nondisabled users rated caption accuracy higher than disabled users did. While 59% of nondisabled users rated the captions as mostly accurate, only 50% of disabled users (including both D/HH and hearing disabled users) gave this rating. Instead, the disabled users were likelier to rate the captions as somewhat accurate (45%) compared to 36% of the nondisabled users.
Participants also indicated whether AI captions are better or worse than human-generated captions. Overall, the sample tended to prefer human captioning, with 64% stating that human-made captions were better than AI-made ones. Another 22% stated they were equally good, and only 14% believed AI captions to be superior to human-made captions. As with the accuracy question above, disabled participants (including both hearing and D/HH disabled participants) had a stronger preference for human-made captions (67%) compared to 60% of nondisabled participants.
When asked if AI captions had made a mistake that negatively impacted them, 84 caption users said yes. This included 13 of the 54 D/HH users (24%), 57 of the 473 hearing disabled users (15%), and 14 of the 315 nondisabled users (4%). Similar to BLV users of AI visual descriptions, D/HH users might use AI captions in higher-stakes situations than hearing users, as the quotes below suggest. Furthermore, D/HH users are also less likely to be able to auditorily verify what was actually said when AI captions make errors. Thus, when AI captions make errors, D/HH users are likeliest to be negatively affected, followed by hearing disabled users.
When examining the mistakes of AI captioning software, participants described several factors contributing to speech misunderstandings, including difficulties understanding accents or unusual speech and difficulties with specialized language. One participant noted weaknesses in the software’s ability to recognize niche language: “I always have to edit YouTube’s captions because it doesn’t understand technical jargon or acronyms. Same with some of the captions I’ve seen on Zoom.” Another participant said, “AI captions are horrible at transcribing speech that is generated by some AAC devices.” A third participant shared difficulties using captions to communicate remotely with a relative: “I use captions to have video calls with my grandmother, who is nearly deaf. It struggles sometimes to accurately convey our conversation, especially if her roommate has the TV on or if a nurse comes. It also does not do a good job of recognizing what she says. I don’t know whether it is because her voice has changed with age or because she has a strong southern accent that is not as common today.” This experience underscores how AI captions may struggle to filter out background sounds and to understand unusual accents.
Participants described negative consequences of AI caption errors, including missing critical information and harming social relationships. One caption user noted that AI captions showed them the wrong date for their citizenship naturalization ceremony. Another participant wrote about a socially harmful mistake, “AI captions misunderstood [a] coworker and incorrectly showed her as saying the N word when she was definitely not saying it, which was distressing to me.” Finally, a participant added, “One skipped word almost killed my dog. Literally.” Thus, errors that occur when AI captioning software mishears speech can have meaningful consequences, especially when disabled users rely on captioning for communication access.
Ideas for Further Innovation in AI Captions
Thirty-two D/HH participants provided suggestions for caption improvements. Within the D/HH group, a major theme was improvements to the functionality of the AI software, such as improved accuracy, the ability to change the pace, and the ability to add context to the captioning tool to improve the accuracy of outputs. One participant indicated, “I would be okay with the transcript appearing more slowly if it meant that the quality was better. Perhaps context should be used to determine the transcription of words that the machine is not immediately clear on.” Another theme related to miscommunication is accents or slang, which also affects accuracy. A participant who indicated a desire for improving how accents or slang are understood wrote, “Ensure that voice recognition testing and training includes people with speech differences. Do not censor captioning, even in media whose audience may include children.” A smaller theme in the D/HH group involved improved accuracy in understanding accents: “Need to make AI able to constantly learn accents, tones, to ensure for accuracy or allow me to correct AI captions for it to learn.”
Similarly, among hearing participants who provided feedback, a major theme was updates to the basic functions of the captioning software, such as fixing mistakes, continuously upgrading the software, and integrating human feedback. One participant indicated, “AI captions could be improved by making them more context-aware, concise, customizable in tone, and accessible for all users.” Some hearing participants also discussed accuracy issues; for example, “AI captions would be better if they were more accurate with timing and punctuation, and if they could recognize different speakers automatically.” The same participant also made a suggestion to improve readability: “It would help to have options to change the caption size, color, and background for easier reading…” The same participant went on to address the need for access to more audible elements when utilizing captioning: “Also, captions should describe important sounds or emotions in the background, not just speech, to make the experience more complete.” A final suggestion was for AI captioning software to be audited for quality. One participant explained their vision for accountability: “Develop or use existing formal metrics to require that AI captions be tested for accuracy against the speech of people with different speech characteristics. Publish the results openly and promote awareness. Disqualify AI caption models that don’t produce equitable results until vendors adequately train their models for equitable output.”
Augmentative and Alternative Communication (AAC)
The third assistive technology explored in this survey was AAC devices and applications. Because understanding gaps in the quality of current AAC can inform current and future AI integrations, AAC users reported on their satisfaction with their current AAC voice and suggestions for improvements. There were 27 participants who reported that they use an AAC device or application for communication. Examples of applications included Proloquo2Go, ElevenLabs, and Speech Assistant AAC. Most participants (70%) said their AAC uses a synthetic voice, while 11% of participants used AAC with a recorded human voice, and the remaining participants were unsure or gave other responses.
AAC voices may vary in quality and comprehensibility. Additionally, some AAC users may feel that none of the voices they can access represent them well. Therefore, we asked AAC users to rate the quality of their primary voice, how well others understand it, and how well the voice represents them. Regarding voice quality, 30% of participants rated it as very good, while 44% rated it as OK, and the remaining 26% rated it as bad or very bad. While 30% of AAC users said people always understand their voice, 52% said people mostly understand it, 11% said that people have an “OK time” understanding it, and 7% indicated that people sometimes or always misunderstand their AAC voice. Just over half of AAC users (52%) stated that their voice is somewhat or very good at representing them, 26% stated it was OK at representing them, and 22% stated that their AAC voice does not represent them at all.
Ideas for Further Innovation in Using AI for AAC
The AAC users were asked for suggestions on how to improve the technology they used for communication. 19 of the AAC users offered suggestions. Many responses emphasized making the AAC voice sound more human-like, with better inflection and emotional differentiation. Some participants also suggested better multilingual support. One participant wrote, “It would work better if the application had faster voice output, more natural-sounding voices, and easier customization options for phrases I use often. It would also help if it could better predict what I want to say based on context.” Another participant reflected on a wish for an AI communication assistant: “I want AI systems like Claude to serve as both my literal voice synthesis and my intellectual advocate, translating my complex ideas into accessible communication while authentically representing my cognitive processing style.”
Discussion
AI is a powerful tool for users with and without disabilities. In this study, substantial numbers of disabled participants used AI to describe images or to caption spoken information. BLV and D/HH participants who used AI for these purposes did so frequently and in a variety of ways. People with other disabilities, as well as some nondisabled participants, also used AI for image descriptions and captions, both to create accessible environments for disabled people and to provide convenience for their own information access.
Although AI holds great promise to improve accessibility of information, the data also show significant gaps between the actual and the ideal accuracy of these AI technologies. Inaccurate image descriptions and captions can have serious life consequences for users, especially those whose disabilities prevent them from being able to access the original images or spoken words to verify accuracy. Additionally, visual description users were concerned about the privacy of sensitive information. They were much less likely to prefer an AI tool for reading sensitive information if it shared the information with a tech company. This was especially true for disabled users, suggesting that disabled users especially wish not to give up their privacy for the sake of access.
Finally, AAC users expressed moderate satisfaction with their communication software, on average. In current and future AAC tools, they prioritized a desire for flexibility in the AAC voice, ensuring that it was human-like and showed appropriate emotional inflection, and emphasized that it should be as customizable as possible.
Recommendations for Innovation
The following recommendations outline actions that those who develop AI for either mainstream or assistive technology purposes should consider to address the opportunities and concerns raised in this research.
Maximize AI development to meet the specific access needs of people with disabilities and enable everyone to use both mainstream and specialized tools effectively.
- Ensure that all platforms that integrate AI in the workplace, including the application process, are fully accessible to and usable by people with disabilities. Adhere to the latest Web Content Accessibility Guidelines and test platforms for use by people with disabilities.
- Prioritize research and development to improve and create tools that specifically benefit people with disabilities or that incorporate the access needs that people with disabilities have.
- Employ people with disabilities as designers, developers, and testers of AI-enabled tools to incorporate more inclusive features into product research and development.
- Incorporate a greater understanding of how people with disabilities rely on or could rely on AI-enabled technologies into product development teams.
- Improve the accuracy of accessibility-related AI uses, such as automated captions and visual descriptions, aiming for higher rates of automatic speech recognition accuracy and more consistent computer vision accuracy. For example, when a blind or low-vision person is taking a picture, ensure the description of the picture is correct and provided in detail.
- Actively train models that detect human speech to detect, understand, and respond to voices with greater diversity, including voice differences related to disability, regional dialects, and accents.
- Wherever AI is used to provide assistive technologies, deploy features that allow users to customize the outputs, such as the size and speed of captions, the verbosity of image descriptions, and the types of voices and extent of emotional inflection in AAC voices.
- Improve the output of AI captions to provide greater information about the environment (including distinguishing between different voices, filtering background noise, distinguishing between speakers, and identifying non-speech sounds) and add features that help users understand the flow of text (such as timing, pauses, and appropriate punctuation).
- Enable user functionality to customize the speed of caption outputs or to slow down the speed in situations where accuracy is paramount.
- Improve the usefulness of image description services by optimizing applications for nonvisual usage and providing tactile or audible cues to ensure that the user is able to easily capture an image for interpretation.
- Provide users with appropriate expectations of model capability to avoid overreliance on AI in highly sensitive contexts, such as transportation safety and visual interpretation of medical information.
- Do not censor content in visual descriptions or captions without the user’s consent.
- Harness AI to create a greater selection of AI voices from which AAC and chatbot users can choose and customize the voices to best represent their personality and use cases.
- Develop AI-enabled AAC features that provide faster and easier means of conveying information and converting it to text output, including developing tools that offer more advanced forms of predictive language output.
- Increase the number of languages in which AI-based captions and AAC are available and improve the ability of the AT to switch between languages.
Improve privacy and data security practices to increase trust in AI products and enable the use of AI with sensitive information.
- Clearly communicate to users how their data is used by the AI developer, in model training, and by third parties, without requiring users to understand complicated terms of service or privacy policies. For example, communicate in the chatbot interface how data is used and any changes to privacy policies.
- Provide users with transparent, easy-to-find controls that allow them to meaningfully decide how their data inputs may be used.
- Provide users with control over how their data is used to train or validate AI models.
- Develop and deploy on-device AI products that allow users to benefit from AI while preserving their data and information on their own device. Allow users to easily switch between cloud and on-device processing or to permanently opt into on-device data processing.
- Ensure data security in high-impact use cases by not integrating sensitive information or conversations into training data by default.
References
Assistive Technology Act, 29 U.S.C. § 3002(a)(4) (2004).
Silverman, A. M., Whistler, A. L., Shock, A., Heydarian, C. H., Baguhn, S. J., Hanuschock, W. E., Hashimoto, M., Khan, O., & Vader, M.-L. (2026). The AI Quagmire: Benefits, Risks, and User Aspirations Through a Disability Lens. American Foundation for the Blind. www.afb.org/AIResearch2
Suggested Citation
Hanuschock, W. E. & Silverman, A. M. (2026). Innovation for Access: AI and Its Use as Assistive Technology for People with Disabilities. American Foundation for the Blind.
IMAGE CREDIT: Select disability-inclusive stock photography courtesy of Disability:IN. Photographs by Jordan Nicholson.