How much work is it to fix website accessibility?

Nearly 1,800 problems. Thirty-four changes. That gap is what this is really about.

A typical site we assessed in the last three months carried, for example, 1,746 recorded accessibility barriers. Almost all of those barriers traced back to 34 distinct changes (in this example): places where a decision in a template, a stylesheet, or a form component was made once. The instance of this decision prevented compliance, and then that template was used to create many other pages.

The point is that one failure ends up being repeated from a single source. These figures come from the government sites we measure, but the mechanism is not particular to government: any site built from shared templates, a private company's included, accumulates barriers the same way. Anyone asking whether to begin addressing digital barriers may be wondering, “Is this a one-off project, or is it an ongoing program?” Thirty-four changes is a project.

What the barriers actually are

The widest pattern, found on 98% of the sites we assessed, is text contrast below the required ratio: text too close in brightness to its background to be read comfortably by people with low vision, or by anyone reading outdoors. It is also the criterion courts have the longest record on, which makes it both the most widespread exposure in this window and the most legally settled.

Custom controls that do not announce themselves (menus, sliders and dialogs built from plain markup with no name, role or state) appeared on 94% of sites. For someone using a screen reader, those elements are announced as nothing: the control exists on the page, but the person cannot use it.

Beyond those two, the pattern holds:

• Form fields without labels appeared on 83% of sites. Fields that rely on placeholder text, or nothing at all, cannot be announced to a screen reader user or navigated by voice control.
• Buttons and controls with weak contrast appeared on 78% of sites: borders and control surfaces too faint to be seen by people with low vision.
• Focus hidden behind other content appeared on 73% of sites. When a keyboard user moves through a page and the focused element disappears behind a sticky header or cookie bar, navigation stops.
• Form fields that do not identify themselves to browsers appeared on 72% of sites, preventing autofill from working for people who depend on it.
• A spoken label that does not match the visible button text appeared on 66% of sites, so saying the button's name aloud does not activate it. Voice control fails at that point.

Text that cannot take extra spacing, a reading aid used by people with dyslexia or low vision, appeared on 84% of sites. Increasing line or letter spacing clips or overlaps the text instead of reflowing it.

Why 34 changes clears almost all of 1,746 barriers

Every page built from one template inherits whatever that template gets wrong. A contrast setting applied once to a heading style propagates to every heading on every page. A form component without a label ships without one everywhere it is placed. The barriers accumulate in the count because they appear on every page; the distinct changes are few because they live at the source.

This is why the barrier count looks larger than the work. On a typical site, five affected templates carry the same inherited failures onto hundreds of pages, producing a large barrier count from a small number of root causes. Addressing the root causes clears almost all of it.

The actual effort to remediate many widespread barriers is rated Medium or Low: scope, not scale. Medium effort means the change needs a developer and a decision, not a rebuild. Low effort, as with text spacing, means a stylesheet adjustment and a test. Neither medium nor low effort barriers should require months to address.

One note on measurement: our detection method changed in July 2026, so figures from before that date and after it are not directly comparable. The findings here reflect late June to late September 2026 and are internally consistent.

Where to start

In cases of color contrast barriers, the case for starting with it is straightforward: it appears on nearly every site in this window, it affects the widest group of people (anyone whose vision is not perfect in every condition), and it sits on settled legal ground. An organization that has not looked at contrast has not looked at its most common barrier.

A practical sequence, working from what the evidence shows:

• Audit the heading and body text styles for contrast against their backgrounds. This is where the 98% figure lives.
• Review every interactive control (buttons, form borders, focus indicators) for the same ratio, which addresses the 78% pattern.
• Check every form field for a persistent, programmatic label that is not placeholder text, addressing the 83% pattern.
• Test keyboard navigation with a sticky header or banner present, to surface the focus-obscuring pattern found on 73% of sites.
• Confirm that custom-built menus and dialogs expose a name, role and state to the browser, addressing the 94% pattern.

None of these steps requires knowing a site's full scope first, and they reduce the barrier count fastest because they address the failures that repeat most widely.

The work is finite. The number to hold onto is 34, not 1,746. For anyone deciding whether to start, that is the number that changes the conversation. See how this maps to your sector (public bodies or private organizations), or how litigation risk maps to these barriers on ADA compliance.