
Short answer
Type "make me a dashboard" into v0 and you'll get one. It will have four KPI cards with numbers nobody gave it, a chart of made-up revenue and a sidebar full of pages that don't exist. That's not a bug. v0 fills every gap you leave with the most likely answer. These prompts close the gaps. If you build in Lovable too, our Lovable prompts follow the same job-by-job layout, written around Lovable's modes instead of v0's features.
First, the name. v0 is Vercel's product, and it now lives at v0.app; the old v0.dev address redirects there. As of November 2026 its pricing page lists Free ($0, $5 in one-time credits for new users, 7 messages a day), Plus ($30 per user a month, shown against a struck-through $90), Business ($100 per user a month, with training opt-out by default) and Enterprise (custom). Plus and Business include $30 of monthly credits per user plus $2 of daily login credits. The features below are the ones v0's docs describe today:
v0's own prompting guide makes the same case these prompts do: start simple, plan complex projects first, be specific, and break big builds into small steps. For a tool-neutral set organized by build phase, see our prompt library for AI app builders.
| Job | Prompts | v0 feature to use |
|---|---|---|
| Agree what you are building | 1, 3 | Plan Mode preset |
| Stop repeating your stack and style | 2 | Instructions |
| Build a landing or waitlist page | 4, 5, 6 | Chat, then a database integration for 5 |
| Build a dashboard | 7, 8, 9 | Chat, then project integrations |
| Build a single component | 10 | Chat with shadcn/ui |
| Turn a design into code | 11, 12 | Attachments or a Figma link |
| Build forms that save data | 13, 14 | Environment variables and a database |
| Polish one element | 15 | Design Mode |
| Change or fix one thing | 16, 17 | Chat, then versions if it goes wrong |
| Get ready to publish | 18 | Security review (paid plans) |
Two minutes here saves an hour of undoing. The Instruction in prompt 2 is the one to keep: it carries your stack, colors and "never invent" rule into every chat.
Turn on the Plan Mode preset first, or paste this as is. v0's docs suggest planning first for anything with several features or user roles.
I want to build [one-sentence description] for [specific user, e.g. dog groomers who run a one-person shop]. The single most important thing a user does in it: [one action]. Do not write code yet. Reply with: the pages and what each one shows, the data you would store (tables and fields), which parts need a database or login, and the 3 questions you most need answered. Keep the first version to what one person can test in a day.
Save this once from the + button in the prompt bar (Instructions, New Instruction) and tick it on for every chat in the project.
Use Next.js with the App Router, TypeScript, Tailwind CSS and shadcn/ui components. Colors: primary [hex], background [hex], text [hex], one accent [hex]. Font: [font name]. Corner radius: [e.g. 8px]. Mobile first: design for a 375px-wide screen, then adapt for larger screens. Use realistic sample content from my brief, never lorem ipsum, invented testimonials, logos or user counts. Every list or table has an empty state, a loading state and an error state. Change only what I ask for. After each change, list the files you edited.
For bigger builds. v0's own prompting guide uses a PRD as the first step for complex projects.
Write a one-page PRD for [product name] based on what we agreed above. Sections: problem, target user, the 3 core user stories, out of scope for version 1, data model, pages, and what "done" looks like for each story. Keep it under 400 words. When I approve it, build only the first user story.Build this with a prompt on Greta.sh — describe the app, Greta writes and deploys it.
A landing page has one job. For deeper conversion work, our landing page prompts cover headline tests and section order in more detail.
One page, one goal. Fill in real copy where you have it; the model will otherwise make up numbers.
Build a landing page for [product], a [category] tool for [audience]. Sections in order: header with logo and a [Start free] button; hero with a headline under 10 words, a one-line subheadline and a product screenshot placeholder; 3 problems the audience has today; how it works in 3 steps; 4 feature cards tied to outcomes; pricing with these plans only: [plan, price, billing period, 4 items each]; FAQ with 5 questions; footer. No testimonials, logos or statistics unless I provide them. Use a labelled placeholder instead.
Pairs a simple page with a real database, so you learn the integration flow early.
Build a waitlist page for [product]: headline, two sentences on what it does, an email field and a [Join the waitlist] button. Store each signup (email, created_at, source from the utm_source query parameter) in a database. If none is connected, ask me to connect one before you write the storage code. Validate the email on the client and the server, block duplicate emails with a friendly message, and show a success state that says [what happens next].
Use when a section works but looks like every other generated page.
Redesign only the [features] section. Keep the copy and the other sections exactly as they are. Give me a layout that is not a three-column icon grid: try [a bento grid / alternating image and text rows / a tabbed panel]. Keep it readable at 375px wide and keep color contrast at least 4.5:1 for body text.
Dashboards are where v0 is most tempted to invent numbers. Give it sample data you control, then wire up the real source once the layout is approved.
Start every dashboard on fake data you control, then connect the real source once the layout is right.
Build a dashboard for [role, e.g. a store owner] to see [what matters, e.g. daily orders and refunds]. Layout: sidebar navigation, a top bar with a date range picker, 4 KPI cards ([metric names]), one line chart for [metric] over time, and a table of the 20 latest [orders]. Generate 90 days of realistic sample data in a separate file so I can swap in a real source later. Each KPI card shows the value, the change against the previous period and a short label. Cards stack to one column on mobile.
The table is where most dashboards break. This one asks for the behavior, not just the look.
Add a table of [records] with columns: [list columns]. It needs: search across [columns], filters for [status and date], sorting on every column, pagination at 25 rows, and a row click that opens a detail panel on the right. Use the shadcn/ui table components. On mobile, show each row as a card with the 3 most important fields. Keep filter and sort state in the URL so a filtered view can be shared.
Run after the layout is approved. v0 can connect databases such as Neon, Supabase and Upstash from the chat or project settings.
Replace the sample data with a [Postgres] database. Create tables for [list], with the fields we used in the sample data plus created_at. Write a seed script with 50 rows so the dashboard is not empty, and keep all credentials in environment variables, never in code. Show me the SQL before you run anything that drops or changes an existing table.
v0 defaults to shadcn/ui, and its docs say Next.js, Tailwind CSS and shadcn/ui give the best results. Go with that unless your codebase uses something else. For how components should look, our UI design prompts cover styles, type and color.
v0's docs recommend building component by component. This prompt forces the states people forget.
Build a [pricing card / file upload box / comment thread] component with shadcn/ui. Props: [list props and types]. Show it in a demo page with every state side by side: default, hover, focus, loading, empty, error and disabled. Keyboard: everything reachable with Tab, Enter and Escape behave as expected, focus ring always visible.
Attach a high-resolution screenshot of your own design or wireframe. Crop tightly if you only want one part.
Recreate the attached screenshot as a working page. Match the layout, spacing and hierarchy closely, but use my colors and font from the saved Instruction rather than the ones in the image. Replace any text that looks like real people, brands or numbers with placeholders in [square brackets]. List anything in the image you could not tell how it should behave, and make the buttons and inputs work.
Paste a frame link for one exact screen, or a file link plus the page name for a full flow. The Figma integration is on paid v0 plans.
Build the [checkout flow] from this Figma file: [Figma link]. The screens are on the page named [page name], frames [frame names]. Use the variables and text styles from the file. Export icons and images from the file rather than drawing new ones. Where the file does not show how screens connect, ask me instead of guessing.
v0 connects databases such as Neon, Supabase and Upstash, plus Vercel Blob, through the Vercel Marketplace, and adding one sets the environment variables for you. Keep every key there, never in a prompt.
Long forms convert better split into steps, and each step is easier for v0 to get right.
Build a [3]-step form for [purpose, e.g. booking a consultation]. Step 1: [fields]. Step 2: [fields]. Step 3: review everything and submit. Show a progress indicator, keep entered values when going back, validate each step before continuing with error messages under each field, and label every field visibly (no placeholder-only labels). On submit, show a loading state, then a success screen that says [what happens next].
Moves a form from demo to real. Keep secrets in the project's environment variables.
Save submissions from the [contact] form to a database table [form_submissions] with every field plus created_at. Then send me an email at [address] with the submission, using [email provider] with the API key read from an environment variable called [EMAIL_API_KEY]. Add basic spam protection: a hidden honeypot field and a limit of [5] submissions per IP address per hour. If sending the email fails, still save the submission and log the error.Want to try one of these? Start on Greta.sh — no setup, no boilerplate — just a prompt.
Iteration is where small, exact prompts pay off. Use Design Mode for visual tweaks to one element, and a chat prompt for anything that touches several. If a change goes wrong, restore the previous version rather than prompting your way back. Building a whole multi-page site rather than an app? Our website design prompts go page by page.
Select the element in Design Mode, then type this in the design panel. v0 attaches a screenshot of the selected element for you.
Make this a three-column grid on desktop and one column on mobile. Match the card style of the [testimonials] section above: same radius, border and shadow. Do not change the text inside the cards.
Use whenever v0 starts redesigning things you liked.
Change only this: [one specific change, e.g. move the pricing toggle above the cards and make the annual option the default]. Do not touch any other component, page, color, font or copy. After the change, list every file you edited and what changed in each.
Paste the exact error. Vague bug reports get vague fixes.
When I [steps to reproduce], I expect [expected result] but get [actual result]. The error is: [paste the error message and stack trace]. Find the root cause before you change code and explain it in two sentences. Then make the smallest fix and tell me how to check it.
Run before you publish. v0 runs a security review only when you ask for one, and on paid plans it always uses v0 credits.
Run a security review. Then check every page at 375px, 768px and 1280px wide and fix anything that overflows or overlaps. Check that every image has alt text, every form field has a label, and every button and link works with the keyboard. Give me a list of what you fixed and anything you left for me to decide.
What is v0?
v0 is Vercel's AI development platform at v0.app. You describe what you want in a chat, and it generates the UI and backend logic as a working app, then publishes it to Vercel. Its docs say it works best with Next.js, React, Tailwind CSS and shadcn/ui. The old v0.dev address now redirects to v0.app.
Is v0 free?
There is a Free plan. As of November 2026, v0's pricing page lists Free at $0 with $5 in one-time credits for new users and a limit of 7 messages a day, Plus at $30 per user a month, Business at $100 per user a month and Enterprise on custom pricing. v0's docs say the older $20 Premium plan is being sunset and is not available to new users.
Which v0 model should I use?
v0 lists four of its own models: v0 Mini, v0 Pro, v0 Max and v0 Max Fast. They are priced per million tokens, from $0.20 input and $1.20 output for v0 Mini up to $10 input and $50 output for v0 Max Fast. Use a cheaper model for small edits and copy changes, and save Max for large, multi-file work. Available models depend on your plan.
Can I use my ChatGPT subscription in v0?
Yes, if you have ChatGPT Plus, Pro or Business. v0's docs say eligible responses from supported OpenAI models then count against your ChatGPT plan, but you still need a v0 credit balance: each request carries a small platform fee in v0 credits, and image generation and some agent work always use v0 credits.
How do I stop repeating the same instructions in every v0 prompt?
Save them as an Instruction. v0's Instructions are reusable prompts stored on your account that you tick on in the prompt bar, and they stay applied until you untick them. v0 also ships two presets: Be Concise and Plan Mode, which writes a plan and asks for approval before changing code.
How are these v0 prompts different from Lovable prompts?
They are written around v0's own features: Instructions, Design Mode, Figma links, screenshots, versions and the shadcn/ui defaults its docs recommend. The job-by-job approach is the same as our Lovable prompts guide, so if you use both tools, the two sets line up.
© 2026 Questera
Follow in Google