What is Meta-prompting?
Meta-prompting asks the model to write or improve the prompt itself — "what would you need to know to do this well?" — before doing the task.
Models are good at knowing what good prompts contain. Instead of guessing the right instructions, describe your goal and ask the model to draft the prompt, list the questions it would need answered, or critique the prompt you wrote. Then run the result.
It's especially useful when you're new to a domain: the model surfaces the constraints and context an expert would ask about, which you'd otherwise discover from a bad first answer.
Example
I want a prompt that produces a weekly investor update from my metrics. Ask me the questions you need answered, then write the prompt with placeholders.
How to use it well
- Ask for the questions first, then the prompt.
- Run the generated prompt and feed the output back for a critique of the prompt.
Go deeper
Related terms
Use it right now
Ask our brain anything on the homepage — it remembers the whole conversation — or write a brief in the Studio and see the prompt it compiles to.