← Library · Core concept

Metaprompting for Self-Correction and Calibration

Sometimes AI outputs miss the mark, not because the instructions were wrong, but because the AI misunderstood the *intent* or *context*. Metaprompting involves giving the AI instructions about how to interpret or execute your subsequent prompts, acting as a 'pre-prompt' for its behavior. For example, before asking for a report, you might state: 'For all following prompts, assume I am a busy executive and require bullet points, concise language, and a summary of key takeaways at the top.' This sets a consistent expectation. Before, every prompt might require re-stating formatting; after, the AI consistently delivers executive-ready summaries.

In plain terms

This is like giving a new assistant a style guide or a set of standing instructions before you give them their first task. It calibrates their approach to all future work.

Why it matters

Metaprompting saves time and ensures consistency by pre-configuring the AI's response style or analytical lens for multiple interactions. Before your next long AI session, try setting up a metaprompt.

Learn one new AI thing every day.

Daily Deck sends you seven plain-English cards like this every morning. Free.

Start free