If a visitor cannot perceive the offer, reach the control or recover from an error, the page has failed before legal compliance enters the discussion.
The 2026 WebAIM Million report automatically tested the home pages of the top one million websites and found detectable WCAG failures on 95.9 percent of them. Automated testing cannot establish full conformance, and WebAIM says so. It can reliably show how frequently basic barriers still reach production. Low-contrast text alone appeared on 83.9 percent of the tested home pages.
Accessibility is often scheduled as a final audit. Performance teams should treat it as part of the conversion path.
A conversion exists only when the action is operable
A campaign report may count the people who completed a form. It does not show the people who never found the label, could not see the focus indicator, triggered the wrong small control or received an error they could not understand.
Those failures are easy to misclassify as weak intent.
The page has several jobs: communicate the offer, establish trust, let the visitor navigate, collect or process a decision, and confirm what happened. Accessibility requirements describe conditions under which more people can complete those jobs. They are not a separate decorative standard.
Contrast changes what can be read
WCAG 2.2 Success Criterion 1.4.3 requires a contrast ratio of at least 4.5:1 for normal text and 3:1 for large text at Level AA, with defined exceptions.
The ratio is a threshold, not an aesthetic instruction. A page can meet it and still create poor hierarchy. A page can also use fashionable grey type that looks refined on a calibrated monitor and disappears on a low-brightness phone outdoors.
Test the states that design files often omit: placeholder and entered text, disabled controls, error messages, links inside paragraphs, hover and focus, selected filters, and text placed on photography.
Do not sample a colour token in isolation. Test the colour in the actual component and state.
Keyboard order is task order
Someone using a keyboard should be able to reach interactive elements in an order that preserves meaning.
A visually neat page can have a broken focus sequence because the document structure does not match the layout. A modal can open while focus stays behind it. A menu can appear without becoming reachable. A sticky consent bar can cover the focused control.
These are performance failures because they interrupt the task.
Start with native HTML controls where possible. They carry established keyboard and accessibility behaviour that a generic container does not acquire when it receives a click handler.
Visible focus is also operational feedback. It tells a keyboard user which control will receive the next action. Removing the browser outline without providing a robust replacement deletes that information.
Accessibility is not a score attached to a page. It is whether the page still works when the visitor does not use it exactly as the designer did.
Target size changes error rate
WCAG 2.2 Success Criterion 2.5.8 sets a Level AA minimum target size of 24 by 24 CSS pixels, subject to specified spacing and other exceptions.
The requirement supports people with reduced fine motor control, but the operational benefit is wider. Small, crowded controls are harder to use on a moving train, with one hand, through a cracked screen or under time pressure.
Do not increase only the visible icon. Increase the interactive target. Check compact navigation, carousel controls, quantity steppers, filter chips, cookie choices and close buttons.
Error recovery belongs in the funnel
A form is not accessible because every field has a label. The user must also understand what failed and how to fix it.
Keep entered data when validation fails. Identify the affected field in text, not colour alone. Move or guide focus appropriately. Explain format requirements before submission when they are predictable. Avoid clearing the entire form after one error.
For payment or legally consequential actions, WCAG includes requirements around error prevention. The design should support review, correction and confirmation rather than treating speed as the only goal. These measures reduce repeated attempts and abandoned tasks for everyone.
Automated checks are the first layer
Automated tools can detect missing labels, certain contrast failures, empty links, invalid attributes and other technical patterns. WebAIM reported an average of 56.1 detected errors per home page in its 2026 sample.
Automation cannot decide whether alternative text communicates the image’s purpose, whether headings explain the page, whether focus order makes sense or whether instructions are cognitively clear.
A useful test stack combines automated scanning, keyboard-only completion, screen-reader checks on critical tasks, zoom and reflow, high-contrast and reduced-motion states, and testing with disabled users where the risk and reach justify it.
Record the task tested, browser, assistive technology and result. “Accessibility checked” is not enough evidence.
The legal date has already arrived
The European Accessibility Act, Directive (EU) 2019/882, applies accessibility requirements to defined products and services including e-commerce, with a main application date of 28 June 2025. We covered scope and evidence separately in what the Accessibility Act actually means for your site.
A WCAG audit is not a complete legal opinion. The practical design response is still clear: know which customer journeys are in scope, which standard you use, who owns remediation and how evidence is maintained.
Do not wait for a complaint to discover that the checkout cannot be completed without a mouse.
Measure accessibility as product quality
Track barriers next to other product defects: critical tasks blocked, issues by severity, components responsible, recurrence after release, remediation time, coverage of manual tests.
Do not reduce the programme to a single automated score. A high score can coexist with one severe barrier on the payment step. Prioritise by user impact and business consequence.
Build accessible patterns into the component system so teams do not solve the same label, focus and error problem on every page. This is the same argument as putting the contrast rule inside a design system that survives a media budget.
Include accessibility in the definition of done for new components. The evidence can be compact: keyboard behaviour, accessible name, contrast combinations, zoom response and the manual test performed.
Test recovery, not only the happy path
Accessible performance includes what happens after a mistake. Submit an incomplete form by keyboard, trigger a payment error and change a selection after validation. Focus should move predictably, the error should be described in text and completed information should remain. A path that works only when every input is perfect is not robust enough for real traffic.
How we handle this
Origin defines accessible type, colour, hierarchy and interaction rules before they become repeated brand patterns. Scale tests the campaign path in the state paid visitors receive, including consent, forms and confirmation.
Merch extends the check through product selection, checkout, payment and post-purchase communication. Commerce accessibility is a chain. Fixing the homepage does not make the transaction accessible.
Send us one critical task, not a request for a score. “Can a visitor choose a product and complete payment?” produces a better audit than “is the site compliant?”
Sources
WebAIM, The WebAIM Million 2026.
W3C, Web Content Accessibility Guidelines 2.2.
W3C, Understanding Contrast (Minimum).
W3C, Understanding Target Size (Minimum).
EUR-Lex, Directive (EU) 2019/882.
European Commission, European Accessibility Act.
Sources checked 30 July 2026. This article describes product practice, not a legal determination.
