AI builders like v0 and Lovable will produce a decent screen from almost any prompt. Getting a good one means telling the tool what a designer would have decided. The 30 UI design prompts below are grouped by section and problem, each with the principle behind it, so you can adapt it rather than paste it blindly.
What makes a good UI design prompt?
A good UI design prompt gives the builder real content, a specific layout decision and constraints it can check its work against. It reads like a short design brief, not a mood. Our guide to product prototyping explains why the content and the question you're testing come before the screen, and the same rule applies to prompts.
Part of the prompt | Weak version | Strong version |
|---|---|---|
Content | “A hero for a SaaS app” | The real headline, subhead and button label pasted in |
Goal | “Make it look modern” | “The visitor should understand the product in five seconds and see one action” |
Layout | “Nice layout” | “Two columns on desktop, text left, product screenshot right, stacked on mobile” |
Constraints | None | “Max width 1200px, one font, three text sizes, spacing in steps of 8px” |
States | None | “Also show the empty, loading and error states” |
Which prompts fix the foundations before any section?
Set the foundations in your first prompt, because every later section inherits them. If you fix type and spacing on screen six, you'll fix it six times.
- “Use one typeface. Use only three text sizes for body, subheadings and headings, plus one display size for the hero.” Why: a short type scale creates hierarchy without noise.
- “Use spacing in multiples of 8px. Put more space between sections than between items inside a section.” Why: proximity tells people what belongs together.
- “Use one accent color only for primary actions and links. Everything else is neutral.” Why: if everything is colorful, nothing signals what to click.
- “Set a max width on paragraphs so body text never runs the full width of the screen.” Why: very long lines are tiring to read and easy to lose your place in.
- “Use shadcn/ui components for buttons, inputs and dialogs, and don't create custom versions.” Why: reusing components keeps the next ten screens consistent.
Which prompts produce a better hero section?
Hero prompts work when they lock the message first and give the layout one job: make the main claim and the main action obvious. Most weak heroes come from asking the tool to invent both the words and the design.
- “Here is the headline, subhead and button text: [paste]. Don't rewrite them. Design the hero around them.” Why: the copy sets line length and hierarchy.
- “Show one primary button and at most one secondary text link.” Why: two equal buttons split attention.
- “Place a real product screenshot to the right of the text, not an abstract illustration.” Why: showing the product answers what it is.
- “Under the button, add one line of proof: [paste a real fact].” Why: proof next to the action reduces doubt at the decision point.
- “On mobile, keep the headline and button above the fold, and move the image below.” Why: the message has to survive the smallest screen.
Which prompts improve feature, pricing and proof sections?
Prompts for content sections work best when they tell the tool how the reader will scan, compare or decide. A feature grid, a pricing table and a testimonial block each serve a different reading pattern.
- “Show three features as rows, each with a short heading, one sentence and a small screenshot. Alternate image side on desktop.” Why: rows with visuals beat a grid of six vague icons.
- “Write feature headings as outcomes the user gets, using this list: [paste].” Why: outcomes are what people scan for.
- “Build a pricing table with [n] plans. Highlight one recommended plan with a border, not a different color for every card.” Why: one emphasis guides the choice.
- “Put the feature comparison in a table below the plan cards, with checkmarks aligned in columns.” Why: tables make comparison effortless.
- “Show testimonials as quotes with name, role and company. Use only these real quotes: [paste].” Why: specific, attributed proof is believable; invented quotes aren't.
Which prompts work for app UI and dashboards?
App UI prompts work when they describe the user's task and the data, not just the screen name. “A dashboard” gives you cards and charts; “a screen where a manager checks which invoices are overdue” gives you a useful layout.
- “The user is [role] and comes here to [task]. Put the information needed for that task first.” Why: task-first layouts beat generic dashboards.
- “Use this sample data with realistic lengths: [paste JSON or rows].” Why: real data exposes truncation and overflow.
- “Show a table with sortable columns, right-aligned numbers and a sticky header.” Why: these are standard rules for scannable data tables.
- “Add an empty state that explains what will appear here and gives one action to start.” Why: an empty screen is many users' first impression.
- “Show loading and error states for this screen, with a retry action on errors.” Why: builders skip states nobody asks for.
For example, suppose a founder rebuilds a settings screen with these prompts
Hypothetical example: a founder asks v0 for “a modern settings page” and gets a long scroll of cards with mixed button styles. They start over. Prompt one sets the foundations: one typeface, three text sizes, 8px spacing and shadcn/ui components only. Prompt two describes the task: “An account owner comes here to change the plan, update billing and invite teammates. Put those three first.” Prompt three pastes real field labels, and prompt four asks for the declined-card error and the no-teammates empty state.
Four short prompts, each with one job, give a screen that's easier to review than one long prompt would.
Which prompts improve forms and onboarding?
Form and onboarding prompts work when they tell the builder what the user must finish now and what can wait.
- “Ask only for the fields needed to create the account: [list]. Move everything else to settings.” Why: every extra field is another reason to stop.
- “Put each label above its field, not inside it as placeholder text.” Why: placeholder labels disappear as soon as someone types.
- “Show validation errors next to the field, in plain words that say how to fix it.” Why: an error far from its cause is easy to miss.
- “Split onboarding into three short steps with a visible progress indicator and a way to go back.” Why: people finish more readily when they can see the end.
- “After sign-up, show one first action that gets the user to something useful, not a tour of every feature.” Why: early value matters more than a full explanation.
Which prompts fix a layout that already looks wrong?
Fixing prompts work when they name the specific problem and the principle, not “make it better”. A vague fix request usually regenerates the whole screen.
- “The page feels cluttered. Remove decorative elements and double the space between sections. Don't change the content.” Why: whitespace fixes clutter more often than new styling does.
- “Everything looks equally important. Make the headline larger and the supporting text lighter, and keep one primary button.” Why: hierarchy needs contrast in size and weight.
- “Text contrast is too low. Make body text meet WCAG AA contrast of at least 4.5:1 against its background.” Why: the WCAG 2.2 AA threshold is a checkable target.
- “Buttons and links look different on each section. Use one primary and one secondary button style across the page.” Why: consistency builds trust.
- “Only change the pricing section. Leave everything else exactly as it is.” Why: scoping a prompt protects the parts that already work.
How do you turn these prompts into a repeatable workflow?
Turn these prompts into a workflow by writing the content first, setting foundations once, then building one section or state per prompt and reviewing each result. It's slower per prompt and much faster overall, because you stop regenerating screens you liked.
Write the words
Draft the headline, section copy, labels and sample data before you open the builder.
Set foundations
Fix type, spacing, color and components in one prompt.
Build one section at a time
Prompt each section separately and accept or reject it before moving on.
Ask for states
Request empty, loading and error states for every screen with data.
Review like a designer
Check hierarchy, contrast, mobile layout and consistency, then fix with scoped prompts.
Prompting gets you a convincing prototype, not a design system or code a team can keep extending. Our comparison of vibe coding vs hiring a team covers the signs you've hit that limit, and shadcn/ui vs a custom design system explains when builder defaults stop being enough. To compare builders, see the best prototyping tools.
When do UI design prompts stop being enough?
UI design prompts stop being enough when the product needs consistency across dozens of screens, real accessibility review and decisions about what to build. Prompts execute design decisions well; they don't make them. If your screens are drifting or users are getting lost, that's a design problem, not a prompting problem.
That's the point to bring in people who design the whole system. Our MVP design service shapes a product's first version, and our product design service covers ongoing work. If you're still testing whether anyone wants it, keep prompting and save the budget.
