Intelligent Accessibility: What If Accessibility Could Adapt to You?

Intelligent accessibility — what if accessibility could adapt to you?

Exploring how AI can transform accessibility from static accommodations into adaptive, human-centered experiences

“The evolution of digital accessibility has given us principles, standards, and tools to identify and remove barriers embedded within digital experiences. But as technology moves towards increasingly dynamic, personalized, and intelligent experiences, can the inclusion of accessibility remain static as it always were till recent? — or more than ever today, do we need to rethink how the term accessibility adapts alongside the people and contexts it is designed for?”

This is a question I have been curious about quite some time. And, it is the remedy that led me to write this article.

Imagine opening a website — You see a navigation bar, a hero image, several buttons, a form to fill out, a few video clips, and a long paragraph of text. Everything appears normal. You know where things are, what they are, and how you are expected to interact with them.

Now, imagine experiencing that same website without being able to see the screen. The interface hasn’t changed, but your ability to interact with it has.

This is where the journey of accessibility begins! It is not with a checklist or a standard, but with a simple question:

How can everyone experience and interact with this in a meaningful way?

Brief history on accessibility?

The term ‘Accessibility’ wasn’t originally a digital design problem.

It grew out of a much broader movement — the recognition that people with disabilities should be able to participate equally in society. Over time, accessibility became deeply connected with ideas of human rights, equality, independence, and participation. It was never simply a matter of making spaces, services, or technologies easier to use; it was also about challenging the social and structural barriers that prevented people with disabilities from participating fully in society.

For much of the 20th century accessibility discussions-movements focused heavily on creating equal oppurtunities for people with disabilities to interact with the physical environment — buildings, transportation, public facilities, workplaces and public services.

Gradually, however, our understanding of disability began to change. This shift did not happen in isolation. It grew through the voices, experiences, and collective actions of people with disabilities and the communities that advocated alongside them.

Instead of seeing disability only as a limitation within an individual, researchers, advocates, designers, and policymakers increasingly began to recognize the role of the environment itself in creating barriers. A person’s experience of disability is shaped not only by their impairment or health condition, but also by the physical, social, technological, and attitudinal environments around them.

This shift introduced a fundamental perspective:

Sometimes, the barrier isn’t the person’s ability. It is how the environment has been designed.

For instance, A wheelchair user isn’t inherently prevented from entering a building because they use a wheelchair; a building without an accessible entrance can create that barrier. Similarly, a blind person isn’t prevented from using a computer simply because they cannot see the screen — the absence of appropriate accessibility support can create the barrier. The WHO uses examples like these to illustrate how inaccessible environments can restrict participation and inclusion.

This way of thinking eventually became central to Digital Accessibility too.

The web has the potential to remove many barriers that exist in the physical world — like distance, mobility constraints, auditory and visual restrictions. But when the web — websites, applications, and digital technologies are poorly designed, they can create new barriers to communication, interaction, and participation.

What is accessibility

There are many definitions on Accessibility, each approaching the concept from slightly different perspectives. Nevertheless, one particularly broad definition comes from ISO 9241 the international standard covering human-system interaction. It says:

“Accessibility is the extent to which products, systems, services, environments and facilities can be used by people from a population with the widest range of user needs, characteristics and capabilities to achieve identified goals in identified contexts of use.”

In simple terms Accessibility means, making sure people with different abilities can access, understand, and use something to achieve what they want to do. It doesn’t specify towards people with disabilities as people can be temporarily disable.

However, Accessibility isn’t only about who is using a product. It is also about, what they are trying to achieve and the circumstances in which they are trying to achieve it — the context. Because a person may be perfectly capable of completing an interaction in one situation and encounter a barrier in another.

For example, Someone who normally relies on audio may suddenly find themselves in an environment where they cannot listen to sound. Someone may be able to comfortably read a screen in a quiet room but struggle to see it in bright sunlight. Someone with full motor control may temporarily have only one hand available. In each of these situations, the person’s underlying ability has not necessarily changed But the — context has.

What is web accessibility

While technology became increasingly integrated into everyday life, accessibility concerns naturally moved into the digital world, resulting the development of one of the most influential foundations in digital accessibility — Web Content Accessibility Guidelines (WCAG). It was first developed by the World Wide Web Consortium (W3C) in 1999, and is being evolving since then.

The W3C Web Accessibility initiative defines:

“Web Accessibility is the principle that websites, tools, and technologies are designed and developed so that people with disabilities can use them equally.”

In practical terms, this means enabling everyone to perceive, understand, navigate, and contribute to the web without barriers.

Furthermore, WCAG has provided recommended four design foundations when testing whether a product product is truly accessible. These principles are named as POUR in short format:

• (P)Perceivable — Information and interface components should be presented in ways users can detect. (E.g., Text alternatives for images or captions for videos)
• (O)Operable — Interface components should be usable through different inputs, including keyboards-only use.
• (U)Understandable — Information and interactions should be clearly understandable and predictable.
• (R)Robust — Content should work reliably across current and future assistive technologies like screen readers and user agents.

Therefore, by adapting these principles into their design practice, designers can begin asking questions such as:

• Can this interface be navigated using a keyboard?
• Can a screen reader understand the structure of this page?
• Does this image have an appropriate text alternative?
• Is there sufficient contrast?

These questions help designers move beyond simply considering accessibility as a checklist and instead build accessibility into the design process from the beginning — keeping a diverse range of people and contexts in mind.

Another important insight is — A website can technically satisfy accessibility requirements and still create a frustrating experience for someone using it.

For instance, a form might have correctly labelled fields, but still contain a confusing sequence of questions. A website might be keyboard accessible, but require dozens of unnecessary key presses to complete a simple task. A video might have captions, but those captions may be inaccurate or poorly synchronized, many more.

Technical compliance and accessibility guidelines can tell us whether certain requirements have been met. But they do not always tell us whether the experience actually works for the person we are designing for.

And that distinction matters.

A product can technically meet accessibility requirements and still be difficult, confusing, or frustrating for someone to use. This is why accessibility cannot be addressed through guidelines and compliance alone. It also requires a connection with real users and an understanding of their experiences, needs, environments, and barriers through user research.

What is intelligent accessibility

Over the years those traditional accessibility frameworks have given us something incredibly valuable — a foundation for removing barriers. But much of these approaches assume that accessibility can be defined through relatively stable characteristics under,

• User needs
• Interface
• Set of accessibility requirements
• Particular context of use

In current process, we identify the barriers target group of users are facing, within those set of characteristics, design against them, and then test whether those barriers have been removed. This approach has been worked remarkably well for many static accessibility problems throughout the history. Therefore, it is not something we should ignore.

Artificial intelligence is changing how we think about accessibility, not by replacing human expertise, but by significantly expanding what is possible. AI can translate, summarize, describe, caption, and reorganize information faster than ever before. As AI reshapes how content is created and consumed, accessible, human-centered design must remain at the core of that evolution.

In AI augmented world, digital interfaces are changing as modern products are increasingly being dynamic, personalized, adaptive, multimodal, and context-aware. (Cao et al., 2025)

https://medium.com/media/22aed53a0cd0595f79f600fbf8b7c397/href

According to CMU’s Digital Accessibility Office views AI as a partner in advancing equity — a way to break down barriers, expand participation, and create multiple pathways to information. AI has the potential to provide humans with greater agency and autonomy, which is the fundamental aim of accessibility.

In systems facilitated by intelligent accessibility infrastructure, content can change based on user behavior. Interfaces can adapt to different screen sizes and respond to diverse user goals and tasks. Recommendation systems can personalize the information we receive, while AI-powered systems can generate content, summarize information, recognize speech, describe images, and modify interactions in real time.

The core idea behind all these approaches is the recognition that people and context are not always static. A person’s needs can change depending on their abilities, environment, activity, goals, and circumstances at a particular moment. What works well for someone in one situation may not necessarily work as well in another.

Who they are > what they are doing > where they are > what technology they are using > and what is happening around them

As we move toward answering the first question I posed at the beginning of this article, the ideas and arguments bring us to the conclusion – digital solutions are increasingly being capable of adapting to how people experience the world. (Hurst et al., 2011)

Therefore, the next step in accessibility is to explore interfaces that can recognize dynamic barriers, understand context, and adapt to people’s changing needs. These interfaces are called ‘Intelligent Interfaces’. It reshape the way we think about accessibility — not by replacing human expertise, but by expanding what is possible with new technologies.

For instance, AI can translate, summarize, describe, caption, reorganize, and interpret information at a scale and speed that was previously difficult to achieve. As AI continues to reshape how content is created, delivered, and consumed, it also creates new possibilities for designing experiences that can respond more dynamically to people and their circumstances.

The term ‘Intelligent Accessibility’ is a way of thinking about accessibility not as something designed into an interface once, asking users to adapt to it, but as something that can continuously respond to the person, their environment, and their context of use.

In this view, accessibility becomes less about providing a fixed set of accommodations to a set of defined users, but more about creating experiences capable of adapting when the conditions of interaction change.

What makes accessibility ‘intelligent’

Traditional accessibility often depends on human-defined rules and predefined solutions for anticipated scenarios. For example, a developer may manually provide alternative text for an image, define semantic structures, or specify how an interface should behave with keyboard navigation.

Whereas, intelligent accessibility introduces another layer: systems that can learn from data, interpret inputs, and adapt their behavior to changing circumstances. Technologies such as machine learning, computer vision, and natural language processing can potentially help systems understand content, recognize patterns, interpret context, and provide assistance that is more responsive to an individual’s needs.

Today, standards such as WCAG provide important guidelines for designing accessible digital experiences, while automated accessibility testing can identify issues such as missing labels, insufficient color contrast, or missing alternative text. However, automated testing has limitations. It cannot fully determine whether an interface is understandable, usable in context, or actually works for a particular person. The UK Intelligence Community Design System, for example, notes that automated testing cannot identify every accessibility issue and that human evaluation remains necessary.

So, what happens in the space between what automated tools can detect and what people actually experience?

Perhaps this is where intelligent technologies can contribute.

Rather than replacing accessibility standards, user research, or human evaluation, AI could potentially complement them by helping systems respond to situations that predefined rules cannot anticipate.

However, this also raises an important question:

“Does making an accessibility system more intelligent necessarily make the experience more accessible?”

The answer cannot simply be yes. It has its own strengths and weaknesses.

Not every accessibility problem can — or should — be solved by embedding AI. Intelligent systems introduce their own technological, ethical, social, privacy, and experience-related challenges. An AI system can misinterpret a situation, make an inappropriate prediction, introduce new barriers, or require access to sensitive personal information.

Therefore, the goal should not be to add AI to accessibility for the sake of making it ‘intelligent’. Instead, the more meaningful question is of us experience designers can ask is:

How can intelligence be used responsibly to make accessibility more adaptive, contextual, and responsive to human needs?

The 4 elements that define ‘intelligence’ in accessibility

In this section, we explore what it means to make a product or a system intelligent from an accessibility. While reviewing foundational research in this domain and seeking to answer the question posed in this subsection, four key elements emerged from the literature that helped shape my understanding of ‘Intelligent’ Accessibility. They are:

These should not be understood as a universally established four-part academic framework. Rather, they are my synthesis of capabilities emerging across research into AI-driven accessibility, adaptive interfaces, assistive technologies, and intelligent environments.

01. Context awareness

Traditional accessibility tools often depend on information that has already been defined by the designer or developer, such as labels, headings, alternative text, semantic structure, and other accessibility metadata.

Intelligent accessibility introduces another layer — understanding the meaning behind what is happening.

For example, a traditional image description might identify an object as ‘a woman’ or ‘a birthday cake.’ A context-aware AI system could potentially go further by interpreting the relationship between these elements and describing the scene as ‘A woman laughing while holding a birthday cake in a dimly lit room.’ The difference is subtle but important: the system is not simply identifying individual elements; it is interpreting them in relation to one another and the surrounding context.

Recent research demonstrates how AI can contribute to this shift. For example, ChartAccessMobile uses computer vision and large language models to analyze charts directly from mobile application interfaces. Rather than simply reading individual text elements, the system identifies charts, extracts their structure and data, generates summaries, and enables users to explore information at different levels of detail. Importantly, the system was evaluated with 10 visually impaired participants, whose feedback highlighted the need to support both high-level understanding and access to specific data within charts (Zhu et al., 2026).

02. Real-time personalization

Most accessibility features today are still largely user-configured. Users may need to enter settings, select larger text, increase contrast, change interaction preferences, activate captions, or choose an assistive technology. These options are valuable, but they generally assume that users already know what they need and are able and willing to configure the system accordingly.

Intelligent accessibility introduces another possibility: the interface itself can respond to changes in the user and their circumstances. Instead of relying entirely on predefined settings, an intelligent system could potentially adapt aspects such as font size, contrast, layout density, interaction methods, or assistance based on changing patterns of interaction and context.

This is where real-time personalization becomes significant. Accessibility can move from a set of options that users configure in advance toward an experience that can continuously adapt as their needs change.

Recent research reflects growing interest in this direction. A 2025 systematic review by Kristić et al. examined 57 studies on machine-learning-based adaptive accessible user interfaces. The review found that adaptive interfaces represented the dominant research direction among the studies examined, while also identifying the importance of balancing automated personalization with user control, transparency, predictability, and the ability to undo or approve adaptations. From experience design perspective, this suggests an important shift — Accessibility can move from something users configure to something the experience can continuously respond to.

03. Pattern recognition

Digital interfaces have traditionally been designed around assumptions of relatively standard interaction patterns,

• A mouse moves in a predictable way
• A finger taps a target
• A keyboard produces recognizable keystrokes
• Speech follows expected pronunciation patterns

But human interaction is much more diverse than these assumptions suggest.

People may interact through eye movements, gestures, speech, facial expressions, alternative controllers, biological signals, or unconventional motor patterns. Even the same person may produce different interaction patterns depending on fatigue, environment, or circumstance.

This is where pattern recognition becomes an important element of Intelligent Accessibility. AI and machine learning are particularly relevant because they can identify patterns across complex forms of input and learn relationships that may be difficult to encode through conventional interaction rules.

Research on AI and digital accessibility demonstrates the use of approaches including machine learning, computer vision, and natural-language processing across a growing range of accessibility applications. A systematic review by Chemnad and Othman (2024), for example, examined 43 studies of AI applications in digital accessibility and identified AI-driven approaches across multiple areas of assistive technology.

04. Predictive automation

The fourth element is perhaps where the idea of Intelligent Accessibility becomes most different from conventional accessibility. Traditional accessibility often responds to a barrier after it has been encountered,

A user encounters a difficult interaction > an assistive technology helps them overcome it

In contrast, intelligent systems anticipate barriers before the user encounters them, following a different sequence,

The system detects patterns or context > anticipates a potential barrier > provides assistance before the barrier becomes a problem

This includes predictive eye-tracking that moves software menus to where a user is looking, or smart wheelchairs that automatically map out and steer around architectural obstacles like steep curbs or stairs. This line of literature defines ‘intelligent systems’ as architectures capable of operating autonomously to reduce human cognitive load. In accessibility, this refers to systems that predict human intent — such as an automated testing suite predicting navigation roadblocks or smart environments anticipating and steering a physical wheelchair around obstacles without user prompting. (de Souza et al., 2023)

These four elements of Intelligent Accessibility — context awareness, real-time personalization, pattern recognition, and predictive automation — are not isolated capabilities; rather, they are interrelated and can work together as part of a continuous adaptive experience.

Context awareness helps the system understand the situation, pattern recognition helps it interpret how a person is interacting, real-time personalization enables it to adapt the experience accordingly, and predictive automation can anticipate what may be needed next. Together, they create a cycle of understanding, interpreting, adapting, and anticipating, rather than treating accessibility as a collection of separate features.

If traditional accessibility focuses on removing known and relatively stable barriers, AI introduces the possibility of responding to barriers as they emerge.

This doesn’t necessarily mean designing AI-powered accessibility only for people with disabilities.

A person can experience an accessibility barrier for many reasons. It could be related to a permanent disability, a temporary injury, the environment they are in, the device they are using, their level of digital literacy, cognitive load, language, or simply the situation they find themselves in at a particular moment.

For example, someone with low vision may need a screen magnified. Someone walking outside in bright sunlight may need the same interface to increase its contrast. A person with a broken arm may temporarily struggle with precise touch interactions. Someone in a noisy environment may need visual alternatives to audio.

Similarly, someone might be tired, distracted, in a noisy environment, using a mall screen, using an unfamiliar device, or dealing with multiple accessibility needs.

Likewise, the underlying needs may be different, but the accessibility challenge is similar — a barrier has emerged between the person and what they are trying to accomplish.

Intelligence does not automatically mean accessibility

It is easy to look at a highly sophisticated large language model or a state-of-the-art computer vision system and assume that greater intelligence automatically means greater inclusion. But embedding intelligence does not automatically facilitate greater accessibility.

AI systems can recognize patterns, generate content, make predictions, and adapt to data. But the quality of those capabilities depends on what data they were trained on, whose experiences are represented, what the system is designed to optimize, and how its outputs are evaluated. An AI system can be highly capable of predicting results and still misunderstand the person using it. Similarly, it can make accurate predictions for one group of users while creating barriers for another. Also, it can make an interface more personalized while simultaneously making it less predictable or less controllable. This is where the human intervention becomes hugely important. Because,

Accessibility is ultimately about the relationship between people, technology, and context. Therefore, when we introduce artificial intelligence into that relationship, we need to consider not only what the system can do, but also how that capability affects the person experiencing it.

It’s a new interesting dimension we are heading towards today.

The challenges of making accessibility intelligent

As we enter this section, I am reminded of a famous quote,

“With great power comes great responsibility.”

In many ways, the same principle applies to the use of artificial intelligence for accessibility — Intelligent Accessibility. As AI gives us increasingly powerful ways understanding context, personalizing experiences, recognizing diverse forms of interactions, and anticipating users’ needs — it also gives us greater responsibility on deciding how that power should be used in real world context with people in need.

This also leaves us with an important question to keep in mind:

“What should AI do, for whom, and under what circumstances?”

This is a question we must consider carefully as creatives working to design for people with diverse needs and circumstances. Therefore, it would not be wrong to say that with greater intelligence comes greater responsibility.

Human oversight remains essential

As AI becomes increasingly capable of generating content, recognizing patterns, making predictions, and suggesting decisions, it can be tempting to ask whether we should let these systems handle accessibility entirely on their own.

In a conversation, this might sound like:

“Now that we have AI tools, systems, and agents, do we really need human accessibility experts anymore?”

From an Experience Design perspective, however, I see it quite differently. Automation should not mean removing humans from the experience — especially when we are dealing with something as sensitive and human-centered as accessibility.

Accessibility has always historically been rooted in understanding people’s lived experiences, needs, barriers, and pain points. These are not always things that can be fully captured by data or predicted by an algorithm.

AI can help us identify patterns, automate repetitive tasks, and expand what is possible. But human expertise, lived experience, empathy, and judgement remain essential in deciding whether an adaptation is actually meaningful for the person experiencing it.

For instance, consider an AI-generated image description. The system might correctly identify the objects in an image, but the description could still miss what is meaningful in that particular context. An automated caption might capture most of a lecture while misrepresenting a critical term. A predictive system might interpret a user’s behavior incorrectly and make an unwanted adaptation.

Therefore, research on AI and disability has similarly identified concerns around bias, discrimination, accessibility, usability, privacy, and the need for interdisciplinary approaches. (El Morr et al., 2024)

Instead of:

AI embedded product/system > Output > User

More responsible model will be:

AI embedded product/system > Human review and contextual evaluation > User

Within this flow, human oversight does not necessarily mean manually checking every AI generated output, instead, it means designing appropriate points of verification, correction, explanation, and user control into the experience. It makes the intelligence accountable to human needs.

Privacy becomes even more important

One of the most promising aspects of Intelligent Accessibility is its ability to adapt to individual needs. But personalization introduces an important ethical question:

“How much personal information should an intelligent system need in order to provide an accessible experience?”

To adapt an interface based on a person’s interaction patterns, a system might process information such as eye movements, speech characteristics, gestures, or other behavioral signals. In some applications, these signals may reveal highly sensitive information about a person’s health or disability.

This creates a difficult trade-off. The same data that allows a system to become more responsive can also increase the user’s exposure to privacy risks.

Research on accessibility datasets has highlighted precisely this tension: disability-related data can help improve inclusive AI, but its sensitive nature also creates risks around disclosure, misuse, and privacy. (Kamikubo et al., 2022)

Therefore, a way out from this sensitive issue would be asking, “What is the minimum information necessary, who controls it, where is it processed, and what happens to it afterwards?”

This points towards principles such as meaningful consent, data minimization, transparent controls, local or edge processing where appropriate, and clear user control over personalization. Personalization needs be an accessibility feature without requirement to surrender privacy.

Avoiding over-reliance on artificial intelligence

There is another risk that becomes increasingly important as AI takes on more responsibilities:

“What happens when the intelligent system stops working?”

If a person becomes dependent on an AI-powered system for navigation, communication, visual interpretation, or another essential activity — a service outage, connectivity problem, model failure, or unexpected system change could have consequences beyond ordinary inconvenience.

This does not mean that AI should not be used in assistive experiences. Rather, the more critical the function, the more important resilience becomes. Especially when designing for people with accessibility needs, intelligent accessibility systems should be designed to fail safely and responsibly. Some of the considerations that can help achieve this include:

• Offline or local functionality
• Non-AI fallback interactions
• Clear error states
• User override and manual controls
• Graceful degradation when intelligence is unavailable
• Transparent communication when the system is uncertain

Of course, these are not the only solutions. Different intelligent systems may involve different contexts, users, risks, and criticality levels, and therefore may require additional safeguards specific to their use case.

The objective of keeping this in mind is not to prevent AI from taking on meaningful responsibilities. Rather, it is to ensure that when intelligent systems fail, the people relying on them are not left without a safe and understandable way forward.

Ultimately, the failure of an intelligent system should not automatically become a human failure.

Conclusion — toward more responsible intelligent accessibility

AI or intelligent systems are neither a magic wand that can eliminate every accessibility barrier nor simply another technology that will disappear with the next trend. Within todays context, it is a powerful set of technologies that can fundamentally change how digital and physical experiences respond to people. Nevertheless, the future of accessibility is not needed to be defined by how intelligent our systems can become.

Perhaps holding hands with the very first question posed in the beginning of this article, we should ask:

“How responsibly can we make them intelligent?”

The answer is given in this piece.

Throughout this article, I explored how accessibility has evolved from addressing relatively stable barriers towards considering the dynamic people, contexts, and circumstances in which interaction takes place. Indeed, still the traditional accessibility principles and standards remain an essential foundation. But as new technologies, interfaces become increasingly dynamic, personalized, multimodal, and context-aware, it’s important to view accessibility in a way that can respond to change rather than simply accommodating it.

The four elements I explored — Context Awareness, Real-Time Personalization, Pattern Recognition, and Predictive Automation — are my synthesis of capabilities emerging across research into topics of AI-driven accessibility, adaptive interfaces, assistive technologies, and intelligent environments. Together, they helped me think about what Intelligent Accessibility could mean when designing for people in needs:

Understand > Adapt > Recognize > Anticipate

But it doesn’t need to be concluded from there.

Each capability also introduces a corresponding responsibility — like,

• Context awareness requires responsible interpretation.
• Personalization requires privacy and user control.
• Pattern recognition requires representative data and inclusive evaluation.
• Predictive automation requires transparency, user agency, and the ability to override.

And most importantly, these systems need to be developed with — not simply for — people with disabilities. The benefits of accessible and adaptive solutions do not necessarily stop with people who have a disability. Because many accessibility features can improve experiences for everyone, depending on the context. It was historically always been that way.

For example, captions were designed as an important accessibility feature for people who are deaf or hard of hearing, but today they are also widely used by people watching videos in noisy environments, quiet public spaces, or situations where they simply cannot use audio. Similarly, voice input, text-to-speech, adjustable text sizes, and other accessibility features can become useful to people for many different reasons.

This is one of the powerful ideas behind inclusive design — when we design to remove barriers for people with particular needs, we may create better and more flexible experiences for everyone.

The principle of ‘Nothing About Us Without Us’ has long emphasized the importance of involving disabled people in decisions that affect their lives.

For Intelligent Accessibility, this becomes even more important. People with disabilities should not be treated merely as the final testing group for an already-developed AI system. Their experiences can help define the problems worth solving, shape the data and interactions used to solve them, evaluate whether adaptations actually work, and determine what responsible assistance should look like.

This brings us back to the role of the Experience Designer — asking questions like,

“What should the intelligence do for this person, in this context — and what should remain within the person’s control?”

Because accessibility is ultimately not about making technology appear intelligent. It is about reducing the barriers between people and what they are trying to accomplish.

• Artificial Intelligence may help us recognize those barriers earlier.
• It may help us respond to them more dynamically.
• It may even help us anticipate them before they occur.

But the intelligence should always serve the person — not the other way around. And perhaps this is where the quote earlier mentioned become relevant once again:

“With great power comes great responsibility.”

As AI gives us greater power to shape how people experience technology. With that power been given, our responsibility as designers becomes greater too.

Intelligence may make technology more capable, but human-centered design is what gives that capability meaning for people and communities.

Perhaps, then, the future of accessibility is not simply about creating technologies that can do more for people, but about creating experiences that can understand, adapt, and assist — without taking away human agency.

And that, to me, is where the idea of Intelligent Accessibility truly begins.

References

Cao, Y., Jiang, P., & Xia, H. (2025). Generative and malleable user interfaces with generative and evolving task-driven data model. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (pp. 1–20). Association for Computing Machinery. https://doi.org/10.1145/3706598.3713285

Carnegie Mellon University. (n.d.). Digital Accessibility Office. https://www.cmu.edu/computing/dao/index.html

Chemnad, K., & Othman, A. (2024). Digital accessibility in the era of artificial intelligence — Bibliometric analysis and systematic review. Frontiers in Artificial Intelligence, 7, 1349668. https://doi.org/10.3389/frai.2024.1349668

de Souza, L. S., Francisco, R., da Rosa Tavares, J. E., & Barbosa, J. L. V. (2023). Intelligent environments and assistive technologies for assisting visually impairment people: A systematic literature review. Research Square. https://doi.org/10.21203/rs.3.rs-3293649/v1

El Morr, C., Kundi, B., Mobeen, F., Taleghani, S., El-Lahib, Y., & Gorman, R. (2024). AI and disability: A systematic scoping review. Health Informatics Journal, 30(3), 1–24. https://doi.org/10.1177/14604582241285743

Hurst, A., Wobbrock, J. O., Gajos, K. Z., Sears, A., Findlater, L., & Trewin, S. (2011). Dynamic accessibility: Detecting and accommodating differences in ability and situation. In CHI EA 2011: Proceedings of the 29th Annual CHI Conference on Human Factors in Computing Systems (pp. 41–44). Association for Computing Machinery. https://doi.org/10.1145/1979742.1979589

Intelligence Community Design System. (n.d.). Limitations of automated testing. Intelligence Community Design System

Kamikubo, R., Wang, L., Marte, C., Mahmood, A., & Kacorri, H. (2022). Data representativeness in accessibility datasets: A meta-analysis. In Proceedings of the 24th International ACM SIGACCESS Conference on Computers and Accessibility (Article 8, pp. 1–15). Association for Computing Machinery. https://doi.org/10.1145/3517428.3544826

Keller, J. (2025, November 12). AI and accessibility: Transforming digital experiences for all users. Accesstive. https://accesstive.com/blog/ai-accessibility-tools/

Kristić, M., Zakarija, I., Škopljanac-Mačina, F., & Car, Ž. (2025). Machine learning for adaptive accessible user interfaces: Overview and applications. Applied Sciences, 15(23), 12538. https://doi.org/10.3390/app152312538

National Public Radio. (2026, March 19). How an act of civil disobedience in 1978 Denver helped propel disability rights. NPR. NPR article

Zhu, X., Shen, M., Zhou, S., Pang, R., Li, L., & Bu, J. (2026). ChartAccessMobile: An intelligent system for accessible chart navigation on mobile applications. In Proceedings of the 23rd International Web for All Conference (pp. 102–112). Association for Computing Machinery. https://doi.org/10.1145/3800424.3800456

---

Intelligent Accessibility: What If Accessibility Could Adapt to You? was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.