Adapting One Landing Page Brief for Six AI Tools
How to adapt one detailed landing page brief for Claude, ChatGPT, v0, Cursor, Lovable and Bolt: what to keep fixed, what to change, and what to check.
The same detailed landing page brief can be used in Claude, ChatGPT, v0, Cursor, Lovable and Bolt, but it should not be pasted into each one unchanged. The design decisions stay fixed: the palette, the fonts, the timings and the reduced-motion path. What changes is the scope each tool works at, where the code ends up and what you check when it comes back.
This guide shows how to adapt one brief for each tool. It is guidance, not a test. HeroPrompts has not run one brief through all six tools and compared the results, so nothing here is a ranking or a finding. For what each tool is best suited to, read the v0, Lovable, Bolt and Cursor comparison. For the general six-question structure, read prompts for AI website builders.
The brief being adapted
The brief below is the one behind Parapet, a SaaS hero for a planning tool. Its opening paragraph names the signature mechanic in one sentence:
"The signature mechanic is an odometer headline: one verb in 'The calm place to ___ your roadmap.' rolls vertically through 'plan', 'write' and 'ship' on a 1.6s loop, while a live roadmap panel below the copy draws its bars in on load."
The colour system is written as named tokens, such as paper: '#F3F0E8', ink: '#1A1916' and signal: '#FF5B3A', and the avoid rule is one line: "No gradients, purple, or glassmorphism anywhere on this page."

What stays fixed in every tool
Some decisions are the designer's to make, and the tool should not make them. These stay identical in every version of the brief:
signal rather than "a tomato red".What changes is the scope, the file instructions and the checks. The sections below set out each one.
Claude in chat
Claude works in a conversation and does not see your project unless you paste it in, so the prompt has to carry everything: the stack, the palette and the file list.
Put the full brief in one message and open with the stack line from the prompt: React 18, TypeScript, Vite, Tailwind CSS 3 and lucide-react. Then ask for the file list before the code: "List the files you will create, then write each one in full." The file structure section at the end of Parapet's prompt is written for this request.
Check that each named token is used by name and that no token has been swapped for a value that looks better. Check that the fonts are loaded once, that every timing matches the spec, that the reduced-motion block is present and that no section appears that the brief did not ask for.
ChatGPT
ChatGPT takes the same chat workflow, so the brief goes in the same way: one message, with the stack line first and the file list last.
The change is in the ending. Repeat the avoid line next to the output request, so the constraint sits in the same message as the instruction the reply answers. If you plan to paste the files back into a project, ask for complete files rather than fragments. A fragment is harder to check against the spec line by line.
Check everything listed for Claude. Then check any copy the tool wrote itself, such as headlines and button labels, against the words you actually want on the page.
v0
v0 builds one component at a time, usually in a Next.js project that uses shadcn/ui. Take only the hero from the brief. The features and pricing sections become separate prompts, each with its own component name and export.
Name the stack and the component library you already have, so v0 does not bring in primitives your project lacks. Keep the palette tokens and font names in the prompt, since the default look is the thing the brief is overriding.
Check that the imports resolve in your project, that the fonts load from the layout your project uses rather than from a file v0 has invented, that the tokens exist in the Tailwind config you are running and that no default card or button styling survived where the brief overrode it.
Cursor
Cursor edits your own repository, so the brief should point at files rather than redeclare them. If the tokens already exist in tailwind.config.js, replace the palette block with "Use the tokens in tailwind.config.js". If they do not exist, make adding them the first step.
Name the file to create or change, and add two constraints: "Do not change any file not listed" and "Add no new dependencies". These matter more in Cursor than in chat, because the tool can reach files you did not mean to change.
Check the diff before you accept it. Confirm that the timing values were kept, that the reduced-motion block was added and that no unrelated file changed.
Lovable
Lovable builds a whole app from a description, including pages, sign-in and data. The hero brief becomes the spec for one page inside that app, so write it as a page: the visual part of the brief, the other pages it sits alongside and the signature moment with its timing.
Keep the hero spec separate from the rest of the app. Later prompts can then change the dashboard or the data model without touching the hero's values. If a form or backend is part of the page, describe it in its own prompt. The Fernhollow contact form is a useful reference for the front-end half of that work.
Check the hero values after the full app has been generated, not only on the first screen, since app-level generation can restyle components. If the tokens have been replaced with the tool's own styling, ask for them back by name. Check that the routes resolve, that the hero works at phone width and that reduced motion shows the finished layout.
Bolt
Bolt is full-stack, so the brief needs to say which parts are marketing and which are backend. Describe the hero on its own, and describe any backend by what it stores or sends. Keep secrets out of the prompt entirely. Name the environment variable a key belongs in, and add the value in the tool's own settings.
Check that the hero still matches the spec after backend work. Backend changes can alter markup or wrap the hero in state that changes its animation, so compare the timings again. Check that no secret has been written into the code.
The checks for every output
Whichever tool you use, the same checks apply, because they test the spec rather than the tool:
If an output fails one of these, the fix is usually a narrower request. Name the line of the spec it should match, say what must not change and ask for one change at a time. Starting over with a longer brief rarely helps. For example: "Change only the odometer roll to 1.6s on a loop. Keep every other value identical, including the token names."
Judge the output against the brief
Judge each output against the brief, not against the other tools. A result that matches the spec is the goal. Whether one tool matches it better than another is a separate question, and this guide does not answer it.
For the wider framework behind this brief, read the six-part prompt framework for web design. For a spec written at three levels of detail, read the same hero section three ways. If you are choosing a tool for the job, browse the SaaS prompts to find a brief close to yours.
FAQ
Can I paste the same prompt into all six tools?
Yes for the design decisions. The scope, the file instructions and the checks should change for each tool, as set out above.
Has HeroPrompts tested its prompts in ChatGPT, Cursor, v0, Lovable and Bolt?
No. HeroPrompts has not run its prompts through those tools and recorded the results. The prompts are written to work in each of them. Every preview on the site is a real build of that exact prompt, made with Claude, so it shows what the prompt produces in Claude.
Which tool should I use for this brief?
Start with the job rather than the brand. Use v0 for a single section you will own, Lovable or Bolt for a whole product, and Cursor for changes to files you already have. The comparison linked above goes into this in more detail.
What do I do when the output does not match the spec?
Narrow the request. Ask for one change, name the line of the spec it should match and say what must stay the same. Check the output against the spec line by line before you ask for the next change.
Build a starter hero prompt in four questions with the free generator, or copy five prompts from the library with no account.
Get the next guide by email
New hero prompts and practical AI design guides like this one, when they go live. No spam, unsubscribe any time.
By signing up you agree to receive these emails. See our privacy policy.
Prompts from this article
Parapet — Modern SaaS Hero
Light, paper-toned SaaS hero with an odometer headline that rolls plan, write, ship, and a live roadmap panel whose bars draw in with a sweeping today marker. Built for product teams and planning tools.
Plinth — Pricing Section with Annual Toggle
A branded three-tier pricing section for a docs platform, with a Monthly/Yearly pill that runs itself on load, a rolling Pro price, and a highlighted plan. Built for SaaS teams that want pricing with personality.
Merrow — Drifting Gradient Background
A pure-CSS six-stop gradient that slides across the viewport on a seamless 18-second loop, with film grain and a vignette. For hero backgrounds, card overlays and section dividers that need atmosphere without a JavaScript loop.
The prompts in the HeroPrompts library are engineered at the level of detail described above — every font, colour, interaction, and animation specified. Skip the iteration and ship a hero section that looks like it cost money.
v0 vs Lovable vs Bolt vs Cursor: Which AI Tool Builds the Best Landing Pages?
An honest comparison of the four leading AI site-building tools for landing pages — what each is actually good at, where each falls short, and how to get the best output from any of them.
GuidesPrompts for AI Website Builders: v0, Lovable, Bolt, Cursor
What v0, Lovable, Bolt and Cursor each need in a website prompt, a six-question brief written for all four, and worked examples from our prompt library.
ToolsThe Best AI Tools for Web Design in 2026
AI web design tools have splintered into distinct categories — chat assistants, visual builders, and code editors. Here's what each actually excels at.

