There's no score, no finish line, and no such thing as a site that's done. That isn't a problem to be solved — it's the nature of the thing. And it's exactly what the checklist mindset gets wrong.
I should say up front this is my read, after a lot of years doing the work. Accessibility has plenty of serious practitioners who'd frame parts of this differently, and they disagree with each other too. Hold onto that — it matters later.
Most people who run a site that fails someone aren't careless. Their site works. They can see it works — they loaded it, clicked through, everything's where it should be. The trouble is they measured usability from their own way of using the web: with sight, with a mouse, moving the way they happen to move. A site that's effortless that way can be a locked door to someone using a screen reader, navigating by keyboard, or working with low vision or none at all. It genuinely passes the operator's test. The operator's test was just never the one that mattered.
Why the checklist feels like enough
Because it confirms what you already believe from looking at your own screen. Run a scanner, check the boxes, call it handled. But the best automated tools catch maybe 20–30% of real issues — 40% on a good day — and only the mechanical ones a machine can reason about. The rest is judgment.
And judgment is contested, which is the part the checklist can't survive. I spent five years on the W3C Education and Outreach Working Group, the people who write the plain-language guidance that practitioners actually use, and what struck me wasn't how settled things were — it was how much was interpretation there is. Competent experts satisfy the same success criterion in different ways and argue about it in good faith. I've got my own positions; plenty of people who do this seriously land somewhere else. That's not a flaw in the field. It's the tell. A thing with that much live disagreement at the expert level was never a box you could finish ticking. There's no "100% compliant" at the end of it — the phrase is a category error. A site is sufficiently compliant or it isn't, and everything past that line is argument and nuance.
Here's the part people get backwards, though, and I had to learn to say it cleanly. They treat compliance as a legal box and usability as the nobler thing floating above it. It's the other way around. If a real person can't use the thing, it isn't accessible — and if it isn't accessible, it isn't compliant. Usability for people with disabilities isn't the higher calling beyond the standards. It's what the standards are for. They're the framework that gets you there. Passing a scanner while a human is locked out isn't compliance without usability. It's a counterfeit of both.
You can watch this go wrong in a specific, common way: the "accessible version." The overlay, the separate stripped-down page, the click here if you use a screen reader side door. It may even clear the automated checks — though nothing ever fully clears them; there's always something flagged. The skill isn't getting to zero. It's knowing which flags matter. A Level AAA item will light up red and may not be worth a minute of your time at the bar you're actually held to; a single unlabeled form field three people hit a day might be the thing that matters most. A scanner can't tell those apart. You have to. And whatever it scores, that side door is still the digital equivalent of sending someone around back to the service entrance — one system for most people and a lesser one for everybody else. Passing, and failing the entire point at the same time. An alternative path to the environment is not equal access. That alone is considered discrimination.
What thirteen years actually taught me
I ran an accessibility consultancy for thirteen years — AKEA, working with global banks, auto manufacturers, and universities, the kind of organizations that get demand letters when this goes wrong. None of that work was simply handing someone a checklist. Most of it was education. Clients arrived scared or angry and, understandably, not knowing much — they wanted the lawsuit to go away, or to be hurt less by the one already moving. The job was quickly alleviating the major pain points to lessen the impact of the legal action. No result buys immunity. If someone wants to sue, they'll find a path — that's the part everyone wishes weren't true.
What actually changes your position isn't a clean scan. It's preparedness: documentation of where you are and what you've fixed, a track record of remediation, a process you can show. The cheapest, sanest version of all this is to build it in from the start instead of retrofitting under duress — planning for it up front costs a fraction of the emergency overhaul, and it signals that the organization took the thing seriously before a letter forced the issue. You don't buy a shield. You build something you can stand behind.
And if accessibility were a checklist, this would be a solved problem by now. It isn't — better than nine in ten sites still aren't compliant, decades into everyone knowing they should be. That number isn't an indictment of anyone's effort. It's evidence about the nature of the work. There's no finishing it, because it's a practice, not an achievement: protocols, habits, the discipline of checking again. The organizations that come through a legal scare in good shape aren't the ones who found the magic green checkmark. They're the ones who made rigor the routine — the steady, documented, unglamorous kind.
How to actually start
Conveniently, you can start this afternoon — no standards body, no consultant. Unplug your mouse and get through your own site on the keyboard alone. Turn on the screen reader for ten minutes; it'll be uncomfortable, and that's the point. Watch one real person use the thing and don't help them — just watch. You'll learn more in ten minutes of that than in ten pages of any audit, and you'll never read a green checkmark the same way.
None of this requires deciding to be a better person. It requires deciding that "passed" was never the bar — because there was never a final test to pass. The work is the bar. The checklist was only ever a snapshot of one moment, seen from one person's way of using the web, on one day. Useful. Nowhere near the whole thing.
