# Structured output · Glossary
Constraining a language model to return data in a defined format, such as JSON that matches a schema, rather than free text.
[Glossary](/glossary) · Agentic systems and generative AI

# Structured output

     Constraining a language model to return data in a defined format, such as JSON that matches a schema, rather than free text. It lets downstream software parse, validate and act on model outputs reliably.

## Why it matters

     Most business uses of language models end with software doing something: creating a record, routing a case or populating a field. Free text is fragile for that purpose; structured output makes the contract between model and application explicit.

     Structure improves reliability but not truth. Values still need validation against business rules and source data, and uncertain or missing values should be representable rather than guessed.

## In practice

     For example, a UK freight forwarder extracting data from commercial invoices might require the model to return a schema with consignee, commodity codes, values and currency, plus a confidence flag per field, so low-confidence values are sent for review.

## Where Rodan fits

     Rodan uses schemas and validation to connect model outputs to operational systems in [AI and Decision Systems](https://rodan.io/what-we-build/ai-decision-systems) delivery.

## Related terms

- [JSON](/glossary#json)

- [Tool calling](/glossary#tool-calling)

- [Intelligent document processing (IDP)](/glossary/intelligent-document-processing)

- [Large language model](/glossary#large-language-model)

- [AI guardrails](/glossary#ai-guardrails)
HTML: https://rodan.io/glossary/structured-output