AI Prompt Builders

Build detailed, model-specific prompts for Claude, ChatGPT, Gemini, Cursor, v0 and Lovable by picking options instead of writing from scratch.

Websites & sections

Builders that turn a website or page-section idea into a detailed, model-ready prompt, with pages, stack, design, SEO and accessibility requirements spelled out.

  • Website Prompt Builder

    Turn your website idea into a complete, model-ready build prompt. Choose pages, stack, design, SEO and accessibility, then copy it for Claude or ChatGPT.

  • Section Prompt Builder

    Generate precise prompts for one UI section at a time — hero, navbar, pricing, cart drawer and 40+ more — with states and accessibility built in.

Popular starting points

Specs, not wishes

Most disappointing AI output starts with a vague prompt. “Make me a website for my clinic” leaves the model to guess the pages, the stack, the colors, the accessibility level and what “done” means, and it usually guesses differently from you. A good prompt reads like a short specification: who the site is for, what each page contains, which technology to use, how it should look and how you will check the result.

Our prompt builders turn that specification into a set of choices. You pick from dropdowns, chips and sliders, start from a preset for your niche if you like, and watch the prompt assemble live. Nothing is sent to an AI from this page. The builders run entirely in your browser, and you copy the finished prompt into whichever tool you use.

Formatted for the model you use

Each AI vendor publishes its own prompting advice, and the builders follow it:

  • Claude receives clearly named XML tags such as <context>, <requirements> and <acceptance_criteria>, with the task at the end. Anthropic’s guidance recommends both.
  • ChatGPT receives Markdown headings, an explicit output format and a scope rule telling it to build only what is specified. OpenAI’s GPT-5 prompting guide recommends this to prevent invented features.
  • Gemini gets the persona, constraints and output format first and the request restated at the end, as Google’s prompt design strategies suggest.
  • Claude Code, Cursor and Codex get a working agreement: inspect the project first, propose a file tree, work component by component and run the tests.
  • v0, Lovable and Bolt get component-by-component instructions with realistic copy and an invitation to ask questions, and can split a big site into a series of smaller prompts.

Detail you control

A detail slider from 1 to 5 controls how much the prompt specifies. Level 1 covers the essentials for a quick experiment. Level 3 adds requirements, constraints and acceptance criteria. Level 5 adds non-goals and an edge-case checklist for serious builds. The builders add structure, not filler, so a longer prompt always contains more decisions.

Share, save and reuse

Every builder can copy the prompt, download it as .txt or .md, and create a share link that restores every choice. Presets you save stay on your device and can be exported as a file for your team. Placeholders like {{brand_name}} can be filled in at the end, so one prompt can serve several clients.