Home/Guides/Meta-Prompting: Asking AI to Write Prompts
Advanced Techniques

Meta-Prompting: Asking AI to Write Prompts

Use AI to design better prompts for itself — a technique that dramatically improves output quality.

7 min read

One of the most counterintuitive but powerful prompt engineering techniques is using AI to write the prompts you'll use with AI. Meta-prompting lets you delegate the prompt crafting to the model itself — especially useful when you know your goal but aren't sure how to structure the instructions to achieve it. It also produces prompt templates you can reuse, share, and iterate on systematically.

What Meta-Prompting Actually Is

Meta-prompting is the practice of using one AI request to generate the instructions for a subsequent AI request. Instead of going directly from goal to output, you insert a design step: describe what you want to accomplish and ask the AI to write the optimal prompt for that task. The result is a more carefully structured prompt than most people write themselves — because the AI has seen millions of examples of prompts and their outcomes in its training data, and can apply patterns that work without you having to know them explicitly. Think of it as outsourcing the prompt engineering craft to someone who has read every prompt engineering guide ever written.

When Meta-Prompting Is Most Valuable

Meta-prompting adds most value in three situations. First: complex or unfamiliar tasks where you're not sure how to structure the instructions. If you're using AI for something you've never tried before, asking it to design the prompt surface things you wouldn't have thought to include. Second: high-stakes prompts that you'll use repeatedly — a customer support response template, a sales outreach formula, a content brief structure. The time invested in good prompt design pays back across dozens of uses. Third: when you have a vague goal and need to sharpen it. The act of asking AI to write a prompt for your goal often clarifies what you actually want, because the AI will ask (or assume) things you hadn't considered.

The Meta-Prompt Template

A reliable meta-prompt template: 'I want to use AI to [describe the task]. The context is [describe who will use this prompt, what platform, what their expertise level is, and any relevant constraints]. Write a detailed prompt I can use for this task. Include: the optimal role assignment, the context the AI needs, the task instructions, any constraints that prevent common failure modes, and the output format. After writing the prompt, briefly explain the key decisions you made and why.' The explanation step is important — it helps you evaluate and improve the generated prompt rather than using it as a black box.

Iterating on Meta-Generated Prompts

A meta-generated prompt is a starting point, not a finished artifact. After generating it, evaluate against your actual use case: does the role make sense? is the context complete? are the constraints appropriate? does the output format match what you need? Make targeted edits, then test the prompt on a real task. If the output falls short, diagnose whether the prompt design was wrong (go back to meta-prompting with more specific requirements) or whether the execution prompt needs refinement (iterate directly on the prompt). Keep the meta-prompt and the resulting tested version — the meta-prompt is your documentation of why the prompt is structured the way it is.

Meta-Prompting for Prompt Libraries

Meta-prompting is particularly powerful for building prompt libraries systematically. For each task type you want to cover: write a meta-prompt describing the task, generate the prompt template, test and refine, then add to your library with a brief description of its optimal use case. This produces a library where every entry has been deliberately designed rather than improvised — which means higher consistency and fewer failures when the templates are used by different team members with different levels of prompt engineering skill.

Prompt examples

✗ Weak prompt
Write a prompt for summarizing documents.

No context about what kind of documents, who will use the summary, what decisions the summary needs to support. The generated prompt will be generic and produce generic summaries.

✓ Strong prompt
I want to create a reusable prompt template for summarizing long-form B2B sales call transcripts. Context: the summaries will be written by sales reps (not analysts) into their CRM after each call. The audience reading the summaries is future reps who might take over the account. Key information needed: buyer's stated pain points, decision criteria they mentioned, objections raised, agreed next steps, and any political/organizational context that affects the deal. Write a detailed prompt template for this task, using [BRACKETS] for variables. After writing it, explain the two or three most important design decisions you made.

Specific task type (sales call transcripts), specific user (sales reps, not analysts), specific reader audience (future reps), explicit required content elements, and asks for design rationale. Produces a template you can deploy immediately.

Practical tips

  • Always ask for the design rationale alongside the generated prompt — understanding why helps you evaluate and improve it.
  • Use meta-prompting when you're entering unfamiliar territory — the AI will surface considerations you wouldn't have thought to include.
  • Treat meta-generated prompts as first drafts — test on real tasks before treating them as stable.
  • Build your prompt library through meta-prompting: systematic design produces more consistent templates than improvised ones.
  • Save your meta-prompts alongside the resulting tested templates — they serve as documentation and make future revisions easier.

Continue learning

Prompt Templates GuideIterative PromptingWhat is Prompt Engineering

PromptIt automatically builds structured prompts for your task — no meta-prompting session required.

PromptIt applies these prompt engineering principles automatically to build better prompts for your specific task.

✦ Try it free

More Advanced Techniques guides

Advanced Role Prompting Techniques

Go beyond 'act as' with layered role prompts that unlock sharper, more

7 min · Read →

How to Build Reusable Prompt Templates

Build a personal prompt library with reusable templates that save time

7 min · Read →

Iterative Prompting: Refine as You Go

Treat prompting as a dialogue — iterate and refine each response to re

7 min · Read →

Self-Consistency Prompting Explained

Improve AI accuracy by generating multiple reasoning paths and selecti

7 min · Read →
← Browse all guides