Use sufficient contrast and readable typography
Text needs enough contrast against its background, and users should be able to resize text without losing information or functionality. Visual design should not depend on subtle colour differences alone.
Make the interface usable without a mouse
Links, menus, forms and controls should work from a keyboard. Focus must be visible and should not become hidden behind sticky headers or overlays.
Give non-text content meaningful alternatives
Informative images need text alternatives that communicate their purpose. Decorative images can generally use empty alt text so they do not create noise for screen-reader users.
Label forms and communicate errors clearly
Placeholders are not a replacement for persistent labels. When validation fails, explain what needs correction and associate the message with the relevant field.
Test reflow and target sizes
WCAG 2.2 includes requirements related to reflow and minimum target size. Responsive design should remain usable when content is enlarged and when users need larger, easier-to-activate controls.
Automated checks are only part of testing
Tools can catch some contrast, markup and labeling problems. Keyboard testing, zoom testing and human review are still necessary because many accessibility issues depend on context.
Accessibility begins in the content model
Clear link wording, logical headings, understandable instructions and meaningful button labels are content decisions as much as technical ones. A developer cannot fully repair an interface whose underlying information is ambiguous.
Include accessibility in acceptance criteria
Specify the expected accessibility standard in the project scope and define how it will be reviewed. Designers can then account for contrast and interaction states while developers build keyboard behaviour and semantics from the beginning.
Use a practical review routine
Navigate templates using only the keyboard, zoom substantially, inspect narrow-width reflow, review form errors, check image alternatives and run an automated scanner. Then manually investigate the findings. The combination catches more than automated or visual review alone.