Today’s outcome

• Inspect desktop, tablet, and mobile behavior
• Combine visual and semantic checks
• Validate AI visual claims
• Describe user impact precisely

Four kinds of visual quality

• composition and layout
• responsive adaptation
• interaction and keyboard behavior
• semantic accessibility

All four affect whether people can use the product.

Visual baseline discipline

Baseline → candidate → controlled environment → diff → human decision

An intentional design change should be approved, not silently accepted or automatically called a bug.

Accessibility is not a scanner score

Check:

• accessible names and structure
• keyboard focus/order
• usable error messages
• contrast and zoom
• meaningful state announcements

Demo: desktop to mobile

At each viewport, test:

1. readable content
2. reachable controls
3. no clipping/overflow
4. expected option state
5. keyboard/touch interaction

AI vision: hypothesis, not verdict

AI can spot possible visual issues.

Confirm with viewport details, DOM/accessibility evidence, keyboard behavior, screenshots, or a human review.

Lab: Digital Showroom Visual Review

Review the launch preview at 1440px, 768px and 375px plus keyboard-only package selection.

Deliver: annotated evidence and user-impact findings.

Strong finding

At 375px width, the launch preview extends beyond the viewport. Record its measured width and affected content; do not infer a nonexistent overlay or focus trap.

Specific user. Specific condition. Observable impact.

Exit ticket

Write one accessibility test that an automated color check would miss.