Lesson 391 of 2116
Structured Outputs: Make the Model Return Data You Can Trust
For production apps, pretty prose is often the wrong output. Learn when to use structured outputs, function calling, and schema validation.
Lesson map
What this lesson covers
Learning path
The main moves in order
- 1Production Apps Want Shapes
- 2Structured Outputs
- 3JSON schema
- 4function calling
Concept cluster
Terms to connect while reading
Section 1
Production Apps Want Shapes
A chatbot can answer in paragraphs. A production workflow often needs fields: title, dueDate, riskLevel, citations, lineItems, actions. Structured Outputs make the model follow a JSON schema so downstream code can trust the shape.
Compare the options
| Need | Pattern |
|---|---|
| Model should call your app's code | Function calling |
| Model should answer with strict JSON | Structured output with text.format |
| You only need valid JSON, not exact schema | JSON mode, but prefer structured outputs when supported |
| User request may be unsafe | Handle refusals explicitly |
A schema is a contract between the model and your application.
const response = await client.responses.create({
model: "gpt-5.5",
input: "Extract a task from: Email Sam by Friday about the vendor renewal.",
text: {
format: {
type: "json_schema",
name: "task",
strict: true,
schema: {
type: "object",
additionalProperties: false,
required: ["assignee", "due", "topic"],
properties: {
assignee: { type: "string" },
due: { type: "string" },
topic: { type: "string" }
}
}
}
}
});- 1Keep schemas small at first.
- 2Require every field you truly need.
- 3Set additionalProperties to false when strictness matters.
- 4Log refusals and parse failures as product signals.
- 5Version schemas when clients depend on them.
Key terms in this lesson
The big idea: structured outputs make model responses easier to integrate, but they are not a substitute for validation, source checking, or user review.
End-of-lesson quiz
Check what stuck
15 questions · Score saves to your progress.
Tutor
Curious about “Structured Outputs: Make the Model Return Data You Can Trust”?
Ask anything about this lesson. I’ll answer using just what you’re reading — short, friendly, grounded.
Progress saved locally in this browser. Sign in to sync across devices.
Related lessons
Keep going
Creators · 9 min
AI Tools: Instructor for Structured Outputs
How Instructor pairs Pydantic models with retries to get reliable JSON from LLMs.
Creators · 9 min
Pro Search vs Default: When To Spend The Compute
Pro Search runs more queries, reads more pages, and routes to a stronger model. It is not always worth the wait — knowing when it is is the skill.
Creators · 10 min
Perplexity API: Building RAG Without Owning The Pipeline
The Perplexity API gives you cited search answers with one call. It is the cheapest way to add grounded retrieval to a product — and the limits are worth understanding.