Teams love one number.
One number feels tidy. It fits on a dashboard. It gives the leaders something to track, the sales team something to sell and product teams something to improve next quarter.
But accessibility does not work that cleanly.
Your product can pass most automated checks and still break for people. That form can have good colour contrast and still announce every error badly. The page scan can come back green and still become unusable when focus jumps around.
That is why I get nervous when accessibility quality gets reduced to a percentage.
That number may be true in a narrow sense. It counts the checks it was designed to count, but it can still miss the thing that matters most. And that's whether someone can use the product to do what they came to do.
A missing image description and a broken keyboard navigation don't even belong in the same bucket.
Accessibility quality is not how many checks passed, but who can still use the product when checks do pass.
That should change what we track and what we measure.
We need numbers that show severity, risk and user impact. We need metrics that point us toward blocked user journeys. I don't care about tidy numbers on pretty-looking reports.
And if AI is going to help with any of that, the question is not whether it can produce a score, but if it can help us see what the score hides.