What is it?
A structured output is a model response that follows a defined schema or format.
The shape can specify fields, types, required values, and allowed options. The application can then validate and use the result more reliably.
04 · Prompting & System Design · Topic 02
What are Structured Outputs?
Structured outputs make an AI response follow a defined data shape.
Think of it like...
Think of it like filling out a form.
The form tells you which fields to complete and what each field means. A schema gives the model a similar shape, but the response still needs validation.
How it works
1 · DefineWrite the schemaSpecify fields, types, and requirements.
2 · RequestSend the taskAsk the model to return the defined shape.
3 · GenerateCreate the resultThe model fills the fields.
4 · ValidateCheck the structureThe system checks types and required values.
5 · HandleContinue or recoverUse the result or handle a validation failure.
Real-world examples
Everyday example · Contact card
Input: a message containing a person's details
AI / System: extracts name, email, and phone fields
Output: a validated contact object
Business example · Invoice extraction
Input: an invoice document
AI / System: returns fields required by the finance system
Output: structured invoice data or an error
What it is NOT
Structured Output ≠ Plain Text
Structured output: follows a defined data shape.
Plain text: has no required machine-readable structure.
Structured Output ≠ Guaranteed Accuracy
Structured output: checks shape and types.
Accuracy: checks whether the values are correct.
Structured Output ≠ Database Storage
Structured output: is the result returned by the model.
Database storage: is what an application may do with that result.
Related concepts
User request → Model response → Schema validation → Application action
Explore next
Remember this
Structured outputs make model responses easier for software to validate and use.
Independent explainer