Google Stitch is positioned to accelerate the start of UI design by getting users past the blank-canvas problem instead of producing finished designs. Clear, detailed prompts produce more relevant AI output. The recommended prompt framework is Zoom-Out-Zoom-In: provide product context and target user, specify the screen goal, layout hierarchy, and design constraints, then state expectations for the generated screen. A concrete prompt can define context, user persona, screen goal, screen type, layout priorities, and individual components such as a sticky header, prominent daily total, horizontal category cards, and a short recent-transactions list. Choosing the right model and constraints yields usable starting designs for human refinement.
"The better you articulate your intention, the more relevant output AI will generate for you. The framework I use for writing prompts is called Zoom-Out-Zoom-In. I start by creating a proper context for my product and explaining its target user, then zoom in on the actual screen/page design I want to generate, explaining the goal of a particular screen/page, its layout hierarchy, and the design constraints the AI should consider when generating it. Finally, I mention expectations I have about the screen that AI will generate for me."
"Context:Design a mobile app screen for a personal finance app that helps users track daily spending.User:Young professionals who check expenses multiple times per day, often on the go.Goal of the screen:Help users quickly understand- Today's total spending- Top spending categories- Whether they are over or under budgetScreen type:Home / Dashboard (iOS-style mobile UI)Layout & hierarchy:- Sticky header with app name and current month- Primary focus: large total spending number for today- Secondary section: horizontal category cards (Food, Transport, Shopping, Other)- Tertiary section: simple list of recent transactions (last 3-5 items)Components:- Header with subtle divider- Large numeric value with supporting la"
Read at Medium
Unable to calculate read time
Collection
[
|
...
]