Pay attention to health tech for even a few weeks and you’ll spot the pattern. Amazon expanded its AI health assistant to anyone on its app or website. Microsoft rolled out Copilot Health, stitching together medical records, wearable data, and personal health history into one AI interface. Anthropic and OpenAI both launched healthcare-focused models for consumers earlier this year. Big tech’s thesis is obvious: AI is the new front door to care.
What none of these announcements mentioned is who gets left outside that door.
More than 70 million adults in the U.S. live with a disability, per the CDC’s latest figures. That’s 28.7% of the adult population. For a huge share of these people, the “front door” tech companies keep building doesn’t open. It doesn’t even have a handle.
Inside an Automated Healthcare Practice: Redesigning Care Around People, Not Paperwork
By reducing administrative burden and redesigning workflows around human needs, it creates space for what matters most: connection between clinicians and patients.
A new layer of disparity
Healthcare has spent decades chasing the promise of better access. Telehealth. Patient portals. Digital intake forms. For many patients, these tools delivered on that promise. But digitization also introduced barriers that fall hardest on people with disabilities.
Think about what it actually looks like. A patient who is blind using a screen reader tries to book an appointment and hits unlabeled buttons. A person with limited hand mobility gets stuck when a chatbot that doesn’t accept keyboard input.
These aren’t hypotheticals I’m inventing to make a point. The CDC reports that 1 in 4 working-age adults with disabilities don’t have a usual healthcare provider. Cost is part of the story, but it’s not the whole story. When the digital systems connecting patients to providers can’t work with assistive technology, they become one more closed door.
The Hidden Administrative Tasks Draining Small Practices
Small practices play a critical role in healthcare delivery, but they cannot continue to absorb ever-increasing administrative demands without consequences.
Now stack AI on top. These new health assistants aren’t side features. They’re being positioned as the primary point of contact for hundreds of millions of users. Microsoft says its consumer products field more than 50 million health questions daily. Amazon’s assistant funnels users straight into clinical pathways, including pay-per-visit channels for 30 common conditions. If those tools aren’t built to work with screen readers, voice input, or switch navigation, you’re not talking about an inconvenience. You’re talking about a new category of health disparity, one where the design of an AI tool quietly determines who gets easy access to a clinician and who doesn’t.
Not malice, neglect
I want to be precise about what’s happening, because the framing matters here. Nobody at Amazon, Microsoft, or any of these companies sat down and decided to exclude patients with disabilities. What happened is simpler and, honestly, more frustrating: accessibility wasn’t treated as a design requirement. It got filed under “later.” Or it didn’t get filed at all.
Health tech has a long track record of this. Screen reader users report that healthcare is still one of the least accessible digital experiences they deal with, right alongside travel booking and mobile apps. Scheduling an appointment, filling out a digital form, finding a provider’s hours, checking whether they accept your insurance. Every one of these tasks can become impossible when a platform hasn’t been tested with assistive technology. Industry audits of healthcare websites consistently turn up hundreds of accessibility violations per page such as unlabeled form fields, missing alt text, color contrast that fails basic thresholds. That’s an inaccessible wall for people with disabilities.
A 2025 research framework in SAGE Journals was blunt about the problem: most AI health tools have been built without meaningful input from people with disabilities. The result is systems that risk producing biased outcomes through poor service quality, misclassification, or outright exclusion. The researchers pointed to a set of gaps that AI could actually help close, including limited access to health information, too few providers who understand disability, and digital tools that shut users with disabilities out. But that only works if accessibility is part of the build, not an afterthought bolted on once someone files a complaint.
The regulatory clock is ticking
Regulators are starting to catch up. HHS has finalized rules requiring healthcare providers receiving federal funding to meet WCAG 2.1 Level AA accessibility standards for patient-facing digital services by May 2026. Websites, patient portals, mobile apps, kiosks. Smaller organizations get until May 2027.
Good. But those rules were written for websites and apps. They weren’t built to account for the speed at which AI interfaces are replacing traditional care touchpoints. When a chatbot is the mechanism patients use to describe symptoms, get triaged, and land in a clinical pathway, the accessibility stakes are different. It’s no longer about whether a website checks the right boxes. It’s about whether someone using a screen reader or voice commands can complete an AI-driven health interaction at all.
For most of these tools, nobody can answer that question yet. And deployment is outpacing testing by a wide margin. For healthcare organizations trying to understand what compliance actually requires, reviewing the HHS final rule and DOJ guidance directly is the most reliable place to start.
What needs to happen
None of this requires a breakthrough. It requires a shift in priorities.
Health systems evaluating AI tools should be asking for accessibility conformance reports before procurement, with the same urgency they bring to HIPAA compliance documentation. Developers building these interfaces need to test with assistive technologies during the build, not six months after launch. And the companies setting the pace here, Amazon, Microsoft, and the rest, should be willing to disclose how their tools perform for users with disabilities. If they’re publishing clinical validation data, accessibility performance belongs in the same conversation.
People with disabilities already face barriers to care that are exhaustively documented. They delay treatment at higher rates. They’re more likely to have unmet medical needs. And they disproportionately struggle with the digital systems that are supposed to help them manage their health on their own. AI genuinely could reduce some of those barriers, by making health information more conversational, more personalized, and easier to act on.
But that only happens if the people building these tools decide, from day one, that accessibility isn’t optional. The industry is pouring billions into making healthcare’s front door smarter. It’s worth asking who still can’t get through it.
Image: Thai Noipho, Getty Images
Mike Barton is a leader at AudioEye, a digital accessibility platform that helps organizations identify and fix accessibility barriers across their web presence. He holds a Certified Professional in Accessibility Core Competencies (CPACC) credential and writes about the intersection of digital access, technology, and health equity.
This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.