January 26, 2026

When accessibility backslides: David Demers' lived experience

By Karine Simard · 5 minute read

At a time when product teams celebrate frictionless journeys and “immersive” interfaces, David Demers sees another reality. As the Executive Director of Accessibility Labs at the Canadian National Institute for the Blind (CNIB), he recounts a life and work trajectory that lays bare a truth: despite solid standards and growing awareness, digital accessibility does not always move forward—and too often, it moves backward.

His story begins long before CNIB. A photographer and head chef, David was sighted. Then a hereditary Leber’s optic neuropathy deprived him of central vision. He retains blurred peripheral shapes, but no more screen reading, no more useful visual cues. The event upended his life. Yet it catalyzed a conviction: technology must be a lever of autonomy, not an invisible barrier.

The first shock, he says, was discovering that the internet wasn’t truly accessible. Irony: pages used to be more readable, because they were simpler, almost purely textual; a stripped-down HTML that screen readers could parse with ease. Today, visual richness has turned into technical complexity: carousels, animations, overlays, modals, weak contrast, text embedded in images. This profusion equates to value for sighted users; for a screen reader, it’s often noise—or even emptiness.

David isn’t speaking in abstractions. He points to everyday tasks—filling out a form, completing a transaction—where the difficulty isn’t intellectual effort, but the absence of access to information. A button without a label, a field not associated with its label, a timer ticking down while a modal window captures the screen reader’s focus, a carousel that advances on its own and erases the content being read. For digital professionals, these “details” belong to micro-interaction design; for him, they are absolute blockers.

He first tried magnification: screen zoom tools, boosted contrast. But when an interface is poorly organized, magnifying isn’t enough. Then he moved to text-to-speech. There, another wall awaited: text placed inside images, buttons designated “button” without explicit role or accessible name, links called “link” without a comprehensible destination. The promise of a universal web stumbles on the prosaic: semantics, focus management, ARIA labels, control names, heading hierarchy, and sufficient contrast.

It’s not just about compliance, David reminds us. WCAG offers a robust framework. When accessibility standards are applied correctly, the experience is smooth, pleasant, sometimes exemplary—even for interactive maps, tagged PDFs, transactional flows. But the field reality is murkier. Organizations often apply the minimum, with little regard for intuitiveness. Partial compliance produces a façade of accessibility: the page “passes” an automated audit, but fails as soon as it’s put in the hands of a blind person.

That’s where his conviction hardens: we need testing with real disabled users. Not at the end of the cycle, not as a token gesture. Integrate lived experience from the design phase, then during implementation, then after each update. Digital professionals know the phenomenon: a redesign breaks a flow, a framework changes a component, a modal library modifies focus handling. Accessibility regressions arrive silently, invisible to the majority. They literally break access for a minority—and the product continues to “work” for everyone… except them.

David sees progress on the government side. In Canada and Quebec, awareness is accelerating, laws are taking shape. The curve is slower for corporations, but he sees a pragmatic understanding emerging: accessibility is good business. The return on investment isn’t limited to screen reader users. Clear interfaces serve SEO, bots, artificial intelligence, and, ultimately, all users. Yet awareness alone isn’t enough. He advocates for firmer legislation, real penalties for companies that don’t meet standards, and a deliberate societal will: make inaccessibility impossible.

His critique isn’t bitter; it’s precise. He highlights product teams’ recurring surprise when they learn he doesn’t follow their site updates. Why would he? Experience has taught him that accessibility isn’t a given. It is won, maintained, then lost at the turn of a version. He calls for clear, proactive communication: inform when changes impact screen readers, announce regressions, indicate patches, share fix dates.

Amid these observations, there are victories. They aren’t grand, but they change lives. Convincing a large company to include tests with blind users, securing the systematic addition of text alternatives for images, enforcing the rule that no button ships without an accessible label—these are “small big victories.” They show a useful shift: compliance isn’t enough; real use takes precedence.

The conversation also turns to artificial intelligence. In the blind community, hope is immense. AI can describe landscapes, interpret photos in context, extract information from a scanned poster, identify interface elements in real time and name them. Behind this potential lies the notion of an “accessibility overlay”: an AI inserted between the user and an inaccessible interface, able to interpret, explain, correct. David tempers: AI itself must remain accessible, inclusive, transparent. It must not repeat the internet’s trajectory, from simplest to least accessible. We should not shift responsibility onto individuals by asking them to install a crutch when the foundation—the product—should already be built for everyone.

To digital professionals, the message is as technical as it is ethical. The technical, first:

Name every control. A button is not a rectangle; it’s an action. Its accessible name must describe that action.
Structure the page. Hierarchical headings, landmarks, tab order, clear and predictable focus.
Avoid text in images; provide relevant alternatives, not generic ones.
Manage focus in modals and dynamic components. Focus return must be explicit.
Respect time. Time limits must be disableable, extendable, or agreed to by the user.
Design for the screen reader first: consider ARIA announcements, roles, states, errors, confirmations.
Test on JAWS, NVDA, VoiceOver, TalkBack. Divergences are a product risk, not a minor QA detail.
Document accessibility rules in design systems and block non-compliant components.
For more info: the practical guide to the digital accessibility lifecycle.

Then the ethical. David speaks of mental load. Sighted colleagues often ignore the invisible effort required to “operate at the same speed.” This effort isn’t heroism; it’s a forced adaptation to avoidable friction. Sometimes a few adjustments suffice: labeling a button, adding an image description, correcting contrast. The cost is marginal; the impact is major. Inaccessibility isn’t an isolated bug; it’s an implicit decision against part of the audience.

What would David do with a magic wand? He would make it fundamentally impossible to create inaccessible content. Programming languages would forbid instantiating a button without an accessible name, CMSs would block publishing an image without a description, frameworks would refuse to compile a component that breaks focus. This vision—“positive security” embedded in tooling—can inspire concrete actions. Teams can configure accessibility lint rules, build UI components that require ARIA attributes and labels, add unit and integration tests for focus and keyboard navigation, integrate automatic checks in CI/CD, and reject merges that degrade accessibility.

Between ideal and practice, the road is clear:

Integrate lived experience from user discovery onward. Include blind people in research, not as an “edge case,” but as a real representation of the audience.
Co-design journeys with explicit accessibility constraints: budget, time, exit criteria aligned to WCAG, but above all to real use.
Equip teams. Design systems with accessibility tokens (contrast, spacing, sizes), accessible-by-default components, documented patterns, embedded checklists.
Test continuously. Automated audits, manual testing, sessions with disabled users at every significant release, and open feedback channels.
Govern. Establish clear responsibilities, measurable objectives, internal penalties for regressions, and—when the law requires—public commitments.
Communicate. Announce changes, publish accessibility notes, indicate known limitations, and provide timelines for fixes.

The Quebec and Canadian landscape is evolving, David observes. The Accessible Canada Act lays a foundation. Governments set the example more quickly; companies follow, sometimes belatedly, sometimes skeptically. But they end up perceiving that accessibility isn’t just a legal obligation: it’s a competitive advantage. Accessible interfaces reach more users, reduce support costs, improve performance and search quality, and prove more robust to technological change.

Beyond metrics, the question is one of meaning. Accessibility isn’t a favor granted to a minority; it’s a design principle. It compels clarity in structure, content, and interactions. What benefits a blind person, by leverage effect, benefits the rest of users: faster navigation, improved understanding, fewer errors, greater satisfaction. The web was born text, it became interface. It must become intelligible again, usable for all.

David concludes without drama. His experience isn’t a manifesto against innovation; it’s a plea for responsible innovation. Digital professionals have in their hands the ability to correct, prevent, and raise the baseline. A rigorously applied standard is useful; an interface tested with blind users is necessary; governance that forbids regression is vital. And AI, if built with them, can become a powerful ally—not a crutch, but an inclusive extension.

The task is vast, but action is possible right now. Name, structure, test, govern. Do not leave for tomorrow what the next version could break. Make inaccessibility impossible, not by magic, but through discipline, method, and respect for people who want, like everyone else, to read, learn, buy, participate. Accessibility is not decreed. It is built, then maintained. And, with will, it can finally progress.

When accessibility backslides: David Demers' lived experience

Further resources