11 June 2025
AI for accessibility and accessibility for AI
AI can expand access to work, community, and opportunity—especially for people with disabilities—by prioritizing accessibility through inclusive design, edge AI, and equitable education.
By Miriam Fernández, CFA, Zoe Roth, Lizzy Moir, and Paul Whitfield
This is a thought leadership report issued by S&P Global. This report does not constitute a rating action, neither was it discussed by a rating committee.
Highlights
AI can enhance individuals' access to society and work by improving the usability of digital content and physical environments, enabling individuals, particularly those with disabilities, to participate more fully in their communities.
Improving the accessibility of AI, and thus its usability, for the widest possible audience will be central to an equitable, AI-centric future and to the technology's potential to improve lives, including by increasing wealth.
Key factors that will drive AI accessibility include: the adoption of edge AI, which will facilitate AI's deployment in physical spaces and objects; investment in AI accessibility; democratization of education with AI; and the embedding of inclusivity and accessibility by design in AI models and solutions.
Louis Braille's invention, in 1824, of the tactile writing system that bears his name was a technological revolution that made the written word directly accessible to millions of people with visual impairments. It resulted in a cascade of benefits for those with limited sight as literacy gains led to improvements in education, which provided access to new employment opportunities, and thus greater economic participation.
Fast forward about 200 years, and the emergence of AI promises a similar revolution for people with a wide range of accessibility challenges, including due to physical, cognitive, educational, and social issues. S&P Global expects that the application of AI technologies to accessibility issues will offer myriad opportunities for greater engagement in physical, intellectual, and economic spaces. Moreover, AI could prove to be a powerful force multiplier with greater participation and technological advancement leading to further, and often likely unforeseen, social and economic opportunities, not only for individuals with disabilities, the elderly, and underserved populations, but also for the broader population.
Yet, innovative AI solutions alone will not ensure that potential benefits are secured or shared equitably including by a broad spectrum of individuals, communities, and economies. There is undeniably also a risk that the technology could exacerbate existing inequalities, particularly with regard to the equitable distribution of wealth and employment.
Access to the technology will require a mixture of robust digital infrastructure and affordable solutions, and it will necessitate that inclusivity sits at the heart of AI systems' design. Making those factors integral to the deployment of AI would help ensure that improvements to autonomy, safety, inclusivity, and economic growth are spread widely, and that AI itself is globally available—a factor we refer to as "accessibility for AI." The likelihood of a positive outcome will be increased by the inclusion of a wide array of stakeholders in the development of AI systems and technologies, and by the deployment of a data-driven decision-making process that prioritizes equality, and notably the needs of people with disabilities and the underserved.
AI and accessibility tools
About 16% of the global population has a significant disability, according to the World Health Organization. AI and AI-enabled tools are already improving the quality of life and access to the physical and digital world for many people. Used by individuals or deployed at a municipal level, the technologies can enhance mobility, offer more personalized services, and reduce barriers for people with disabilities, the elderly, and the underserved.
Shared environments
In shared environments, AI-driven solutions can form a digital layer that promotes equity, particularly where they help make public spaces, services, and civic participation more inclusive by design. Some of the key tools being deployed are:
Computer vision (CV): Wearables (such as eyeglasses) and public CV systems deployed at intersections or on streetlights can identify obstacles in real time and alert users, enabling smart navigation for people with low or no vision or mobility challenges.
Internet of Things (IoT) sensors: Connected sensors embedded in public infrastructure can improve safety and comfort by monitoring air quality, lighting, or curbside conditions. For example, citizens with respiratory issues can be alerted to remain inside or avoid certain areas when sensors detect that pollution has passed a predetermined threshold.
Predictive analytics: When embedded in user-facing applications, this technology can help anticipate transit demand or service outages. Those predictions can facilitate planning and enable proactive communication to assist people with visual or mobility impairments.
Generative multimodal AI: This technology's ability to create new content (including text, images, audio, and video) can be integrated into municipal websites to translate complex, multi-source, and multi-format data into accessible formats. Audio, simplified visuals, and multilingual summaries can be tailored to users' needs and preferences.
Private residences
In the home, AI-driven technologies can significantly enhance safety, independence, and quality of life, particularly where they are combined to form an intelligent support system using sensors, automation, and predictive capabilities that can assist with daily living and risk monitoring, and promote autonomy. Technologies being used in homes include:
Computer vision: CV-enabled devices such as fall-detection cameras or wearables and smart glasses use visual AI to recognize accidents, monitor daily routines, and assist people with visual impairments by identifying objects or reading text aloud.
Voice assistants: Hands-free tools such as Amazon Echo or Google Assistant allow users to control lighting through integration with smart home systems, call for help, or receive reminders. Voice control makes technology more accessible to people with mobility or dexterity challenges.
Smart sensors and IoT devices: Motion detectors and connected appliances gather data from the home environment, automatically triggering actions such as shutting off a stove or alerting caregivers when routines are disrupted.
Predictive analytics: These systems, either through integration in wearables or through sensors, analyze behavioral patterns to provide early health risk warnings. They can anticipate falls, detect changes in mobility, and send medication adherence alerts before problems escalate.
The application of AI technologies to enhance accessibility is already widespread, as the following (non-exhaustive) table of examples demonstrates (see table 1).
Technology is evolving to support accessibility
Technological advancements and a shift toward accessibility-first design are changing the traditional assistive technology (AT) market. For example, wheelchairs and hearing devices have matured over the past two decades thanks to advancements in technology and manufacturers' increased commitment to inclusivity and accessibility. Principles around operability, understanding, robustness, and perception of the physical environment have gained prominence in designing accessible environments. More recently, the integration of IoT devices, including wearables; telemetry, including remote monitoring and emergency response devices; and multimodal generative AI, which facilitates natural language communication with technology, has enabled more holistic technological support that has contributed to easing cognitive, digital, and physical barriers for individuals with disabilities.
Looking ahead, agentic AI promises autonomous execution of tasks, memory-enabled AI, greater context understanding, and reinforced learning. That, coupled with edge devices that will have an increased capacity to compute more complex AI algorithms, will further alleviate accessibility barriers that currently restrict disabled individuals' access to the labor market. On average, persons with disabilities present an unemployment rate 4.7x higher than the total unemployment rate, according to data drawn from European countries and the US over a five-year period (see figure 1).
AI can help address this issue, thanks to its recursive nature, meaning it uses its own results as input for self-improvement. This characteristic of AI is particularly suited to the enhancement of accessibility processes that aim to improve access to physical and digital environments. For example, the Web Content Accessibility Guidelines (WCAG) aim to bridge the gap between available digital content and the usability of that content for individuals with disabilities. With AI uncovering previously hidden gaps in usability and proposing solutions to improve website design, it will be easier to execute, automate and scale digital environments for individuals with disabilities.
Likewise, we expect usability of digital content will improve thanks to the soon to be applied European Accessibility Act (EAA), which is likely to catalyze innovation in inclusive design and the development of accessible environments. For instance, AI-powered ATMs in Europe are required to offer content and layouts calibrated to people’s age, financial literacy, and visual or cognitive disabilities. In practice, this necessitates offering simplified menus and adjusting font size and screen contrast, among other possible adaptations.
AI's impact will partly rely on its accessibility
AI has the potential to be broadly beneficial to humanity, but to facilitate that outcome, the technology will have to be widely accessible. Efforts to expand AI's usability will thus be a key determinant of AI's equitability and will partly determine the extent to which it contributes to improvements in income per capita across societies.
Based on our analysis conducted in April 2025 using data from 2020 to 2024, Singapore, Saudi Arabia, Andorra, Estonia, Ireland, and Finland appear best positioned to support AI accessibility for their populations (see figure 2). Our analysis included four factors that we consider key to ensuring accessibility to AI:
Internet cost as a percentage of gross national income per capita
The percentage of households with internet access
Education levels, as measured by the out of school rate for people of upper secondary school age
ICT graduates, as measured by the percentage of graduates from information and communication technologies (ICT) programs.
Frontier markets are susceptible to AI accessibility issues
Our analysis suggests that frontier markets are likely to experience the greatest challenge in terms of providing AI accessibility to their populations. Frontier markets are defined as countries with average annual per capita income below $2,500 (on a GDP basis), which face greater-than-average economic challenges and financing needs, and which S&P Global Ratings has assigned a sovereign credit rating of 'B' or lower. Frontier countries are home to about 17% of the world's population, while emerging market countries house about 61%.
Based on our accessibility for AI analysis, Madagascar, Benin, Rwanda, Cameroon, and Honduras appear least equipped to provide the infrastructure and resources to enable their populations to effectively access AI. The main constraints for frontier (and some emerging) countries are limited access to the internet and a lack of technological skills, as measured by the number of ICT graduates in their workforces (see figure 3). Those constraints are likely to lead to a broadening of the inequality gap, both within and between economies. We believe that closing that gap will require coordinated actions, including by national governments, the private sector, and multilateral institutions. One such initiative is the investment by Microsoft and UAE-based G42 to build Kenya’s AI infrastructure, including a data center to develop local-language AI models and research.
What comes next
We have identified four key elements that are likely to play a critical role in promoting AI accessibility and thus reducing the risk of persistent or growing inequality, both within nations and across the globe.
Edge AI
Edge AI refers to the deployment of AI algorithms in local (or edge) devices, including physically driven accessibility infrastructure (e.g. smart homes or wearable health monitors), such that they can function without relying on constant internet connectivity. This is critical for underserved areas including rural towns, low-income communities, and some frontier and emerging economies where access to reliable high-speed internet service is not always available. By processing data at the edge (on devices or at local gateways), AI-powered accessibility tools (including computer vision for obstacle detection, natural language interfaces, and mobility assistance devices) can be deployed more widely and equitably.
Democratization of education
AI has the capacity to transform the education industry by expanding access to affordable, high-quality teaching, and by providing education resources to support learning among lower-income students. The resultant potential for AI to reduce the role of wealth as a determining factor of educational quality could improve educational outcomes, ranging from basic literacy to access to advanced education, in lower-income economies and thus help drive economic growth.
Investment in AI accessibility
One of the major challenges facing the development of AI accessibility solutions is a lack of research and development, due to the paucity of funding for such initiatives by both the private and public sectors. The cost of accessing the data and the computing power needed to develop AI accessibility has constrained new projects, as have the salaries and fees required by talented AI developers, the OECD noted in its paper, "Using AI To Support People With Disability In The Labour Market," published in November 2023. We expect the ongoing shift to open-source advanced-reasoning generative AI models will make such investments more affordable because individuals won´t need to train these models from scratch, but will be able to use and fine-tune established models to meet their needs. This will also alleviate some of the financial pressures of development for those with limited means.
Embedding inclusivity and accessibility by design
There is a risk that AI models and AI solutions will perform unequally across different user groups, including due to inbuilt bias or design that is exclusive of people with disabilities. The potential for such embedded exclusion and resulting user frustration poses an argument for AI models to be designed with inclusivity as a goal. Accessibility and inclusivity are generally governed under broader anti-discrimination regulations, such as the Equality Act in the UK, the law on the inclusion of persons with disabilities in Brazil, and the Disability Discrimination Act in Australia. Ensuring the enforcement of inclusivity in AI models can be challenging due to the often-inherent lack of transparency of training data and algorithms. However, opportunities to evaluate AI solutions that are directly available to customers will likely make it easier to test for inclusivity and accessibility.
Content Type
Research Council Theme