AI Glossary Term
What does “Structured Output” mean?
Definition
Model output constrained to a defined schema such as JSON, tool arguments, or typed fields.
How Structured Output works
Example
An expense app asks the model to return a fixed set of fields: vendor, date, and amount. Because the reply follows a schema, the app can store it directly without guessing how to read a sentence.
Why it matters
Reliable structure lets AI plug into software. It reduces parsing errors, though the values inside can still be wrong.
Common questions
How is it enforced?
By providing a schema and using model features or decoding constraints that limit output to it.
Does it make values accurate?
No. It guarantees shape, not truth, so values still need validation.
Structured output vs function calling?
Function calling is a way to return structured arguments to trigger actions. Both rely on schemas.