How to test an AI-generated UI patch before trusting the screenshot

A screenshot can make an AI-generated UI patch look finished even when keyboard flow, mobile layout, loading states, or long text are broken. This checklist shows what to test before accepting the patch.

## Check interaction, not just appearance
Click the main controls, tab through the page, submit empty forms, and try long labels. Many generated patches optimize the first screenshot but miss focus states and error states.

## Use at least two widths
Desktop-only screenshots hide text wrapping and crowded controls. Test a narrow mobile width and a normal desktop width. If a toolbar, card, or modal changes height unexpectedly, the layout is not stable yet.

## Look for data edge cases
Try missing images, empty lists, long names, and slow loading. A UI patch that only works with perfect sample data is still unfinished.

## Accept with evidence
The useful handoff is not “looks good.” It is a short note saying which viewport, interaction, and edge case were checked. That evidence is what makes the screenshot trustworthy.

How to test an AI-generated UI patch before trusting the screenshot

// COMMENTS

ON THIS PAGE