Band-Aid Accessibility
Disabled users are increasingly relying on LLM-powered coding tools to patch inaccessible software that mainstream companies continue to ignore.
Deep dive
- Disabled users are turning to LLMs to build accessible interfaces that mainstream software lacks.
- Creating these workarounds leads to 'band-aid' accessibility that may not be sustainable long-term.
- There is a risk that companies will use the existence of these community-led patches to excuse their own failure to design inclusive products.
- The practice creates parallel, disconnected software spaces rather than integrating accessibility into standard development workflows.
- The author argues that true accessibility requires hiring specialists and integrating inclusive design from the start, rather than relying on LLM-generated patches.
Decoder
- Screen reader: Assistive technology software that reads out the content of a computer screen for visually impaired users.
- Accessibility-wash: The practice of superficial or performative efforts toward accessibility to improve corporate image without making substantive changes.
- WCAG: Web Content Accessibility Guidelines, the industry standard set of recommendations for making web content more accessible.
Original article
More and more disabled people take matters into their own hand and use LLM coding tools to generate accessible features or wholly accessible app for their use cases. While it seems easy to celebrate this as an act of liberation for these users, reality is more complicated.
Preamble
I support everyone to get the access needed, even if this is via dubious, unethical, or illegal ways. I can't fault someone with chronic pain who finds relief in cannabis. I won't judge anyone who needs to find "alternative ways" to get books that work for them and their disabilities, and who cannot read DRM-laden books. And I don't find fault with anyone who removes barriers using LLM coding tools.
This is an absolute stance.
Why this happens
There is a lot of debate, mainly on Mastodon, about the ethics of creating(?) accessible apps this way. And a lot of these arguments, I can relate to:
- Disabled people have always been ignored by most mainstream software developers and companies.
- Years of begging to make interfaces accessible, particularly to screen readers and for their users, lead to nothing.
- This is a way users can ensure access in a timely way.
- Screen reader users know better how interfaces work for them and can "prompt" to meet these requirements.
- The projects might not be sustainable in the long run.
- Not using mainstream software creates parallel spaces for disabled people.
- Software developers might do less accessibility, because disabled people can just patch it after the fact.
- Commercial LLM vendors use these projects to accessibility-wash their ethical problems.
The state of accessibility is dire. I review dozens of projects every year, and the quality has rarely been as low – with notable exceptions! – than in 2026. And this doesn't even cover only accessibility issues: It looks like most software is not tested at all, be it for functionality or error handling, or just stability. Everything breaks all the time.
And what is an inconvenience to some users is a barrier to others.
Conclusion?
As a LLM skeptic, I see a lot of the risks outlined above myself. It would never be for me to use an LLM to create something. I enjoy the creating and the problem solving too much, and honestly as an accessibility reviewer, you often get the "identifying problems" but hardly ever the "problem solving" side of the coin. It can be very frustrating.
But this is for many screen reader users the first time they feel that they have tools they can use to solve accessibility problems by themselves for themselves. I always wondered why there are not more disabled, and especially blind, programmers around, and I assume the reasons are ableist structures around computer education.
But even then, letting a blind developer lead your website or app UI development? Probably unimaginable in most companies. UI means painting it nicely, right?
My real worry is about the long-term sustainability of this approach. Creating separate paths often means features get out of sync. Rising cost of LLM code generators will also limit what can be done on a budget. Which also makes these kinds of developments, as accessible as the result may be, burden with a disability tax. And that might even increase once companies de-prioritize accessibility work because supposedly the community will take care of it.
What I hope this does is for disabled people to be interested in coding more, but we also need the other side to go and do the work: It can't be that accessibility issues are pushed later for years and years and years.
No doubt, the community is creating some extremely good band-aids, but it is embarrassing that they are even needed.
If you want to help to remove the reliance on LLM coding tools by disabled people, you have to make your stuff accessible, and listen to their feedback. Or even better: Hire that blind UI specialist.
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi