I’m using AI a lot more than I did a few months ago because I’m currently the primary caregiver for both of my parents.
That means my time, attention, and energy are fragmented in ways that are hard to explain to people who haven’t lived it. I can still think. I can still create. I can still do meaningful work. But I can’t always do those things on a predictable schedule or move through every step of a task without interruption.
AI helps me bridge that gap.
It helps me hold onto a thought when I’m pulled away. It helps me organize information when my brain is carrying too much. It helps me restart after an interruption without having to reconstruct everything from the beginning. Sometimes it turns a task that feels impossible in the moment into one I can actually finish.
That is not laziness. It is not cheating. It is an accommodation.
And that is where conversations about AI often reveal an ableist assumption: that “real” work must be completed independently, through sustained concentration, in a particular sequence, and without assistance. The person who can sit uninterrupted for hours is treated as more disciplined or more authentic than the person who uses tools to work around pain, fatigue, executive dysfunction, cognitive overload, or caregiving demands.
But independence has always been partly fictional. People rely on editors, assistants, search engines, templates, colleagues, medications, mobility devices, screen readers, calendars, and countless other forms of support. We tend to call a tool legitimate once it becomes familiar—or once the people using it have enough status.
The accessibility possibilities of AI extend far beyond caregiving.
AI-powered speech-to-text can help people who cannot comfortably type, whether because of arthritis, tremors, paralysis, chronic pain, repetitive-strain injuries, or limited fine-motor control. Voice commands can help someone navigate a computer when using a mouse is difficult or impossible. Text-to-speech can support people with low vision, blindness, dyslexia, migraines, fatigue, or other conditions that make reading from a screen difficult.
Live captions and transcripts can help Deaf and hard-of-hearing people, people with auditory-processing disabilities, and anyone who cannot reliably process spoken information in real time. Image descriptions can make visual material more accessible. Predictive text and communication tools can help people with speech disabilities express themselves more quickly.
AI can also reduce cognitive load. It can break a large task into smaller steps, summarize a dense document, reorganize scattered notes, explain unfamiliar language, or turn a disordered first draft into something easier to revise. Those functions may help people with ADHD, dyslexia, brain injuries, memory impairments, chronic fatigue, autism, or cognitive disabilities—but they can also help anyone whose capacity changes from day to day.
We see the same ableism in reactions to ordinary convenience products. People mock pre-peeled oranges, pre-cut fruit, peeled hard-boiled eggs, and other prepared foods as symbols of laziness or absurd consumerism. There are legitimate questions about packaging, cost, and waste. But the ridicule often assumes that everyone can peel an orange, cut vegetables, or remove an eggshell easily and without pain.
For someone with arthritis, tremors, limited dexterity, chronic pain, paralysis, fatigue, or another disability, prepared food can be the difference between eating independently and not eating at all. What looks unnecessary to one person can be life-changing to another.
The same is true of energy.
“Spoon theory,” a metaphor created by Christine Miserandino, describes the limited units of energy available to many people with chronic illnesses or disabilities. Every task costs a certain number of “spoons,” and some people begin each day with far fewer spoons than others. Showering, getting dressed, making a phone call, preparing food, or answering an email may use energy that cannot simply be replenished by trying harder.
A tool that saves several steps can therefore preserve enough capacity for something else: eating dinner, attending an appointment, caring for another person, completing paid work, or simply making it through the day.
AI can function in a similar way. Drafting a first paragraph, organizing scattered thoughts, transcribing speech, or helping someone resume an interrupted task may conserve a few spoons. That does not mean the person contributed nothing. It means they used a tool to decide where their limited energy would matter most.
Spellcheck is a useful example. It can be indispensable for people with dyslexia and other language-processing disabilities, yet it is now so ordinary that most people no longer consider using it dishonest. We do not insist that every correctly spelled word must prove the writer could have produced it without assistance. We understand that the writer’s ideas, choices, and judgment still matter.
Many technologies begin as accommodations, are dismissed as shortcuts, and eventually become ordinary tools. The social judgment often fades once nondisabled people find the technology convenient.
That should make us question what we mean when we say AI makes something “too easy.” Easy for whom? Compared with what? And why is difficulty treated as proof of merit?
AI deserves scrutiny. These systems are trained on data created, collected, and selected within societies already shaped by racism, sexism, ableism, classism, and other forms of inequality. AI can absorb those patterns and reproduce them in its language, assumptions, recommendations, and omissions. It can treat the experiences of dominant groups as the default while misunderstanding, stereotyping, or excluding everyone else.
Those biases can become especially dangerous when AI is used to make or influence decisions about employment, healthcare, education, policing, insurance, or access to public benefits. A biased system does not become objective simply because its discrimination is expressed through an algorithm.
The harms can also overlap. A disabled woman of color, for example, may encounter racism, sexism, and ableism at the same time. If the people testing an AI system do not include people with those intersecting experiences, serious problems may remain invisible until the system has already harmed someone.
AI also raises questions about labor, privacy, environmental cost, accuracy, consent, and power. Disabled people may benefit enormously from these tools while also being among those most harmed when automated systems are carelessly designed or used to make decisions about their lives.
These systems are imperfect even when discrimination is not the issue. Transcription can mishear people, especially those with speech differences or accents. Image descriptions can omit what matters. Summaries can erase nuance. AI can confidently produce false information. Accessibility technology must be designed and evaluated with disabled people—not merely marketed to them.
None of that means we should reject AI’s value as an assistive tool. It means accessibility and accountability must develop together. We should not have to choose between gaining access to helpful technology and being protected from the harm that technology can cause.
The question cannot simply be whether someone used AI. We should ask what they used it for, what judgment they retained, whose work or data made the system possible, what risks were created, and whether the tool expanded someone’s ability to participate.
For me, right now, it does.
It does not replace my judgment, my voice, or my responsibility for what I produce. It helps me continue to use them during a period when caregiving has changed what my days—and my capacity—look like.
People who have never had to count their spoons may mistake assistance for an unfair advantage. Often, it makes participation possible—not effortless.
That distinction matters.