All prompts

Structured output prompts

Prompts that pin down the shape of the answer, not just its subject.

When output feeds something else — a spreadsheet, a parser, a document template — the format is the requirement. These prompts declare the schema, the field types, what to emit when a value is unknown, and whether prose is allowed around the payload at all. That last one is what usually breaks an integration.

3 prompts published

Guides on structured output prompts

Questions about structured output prompts

What does a structured prompt pin down?
The schema and its field names, the type of each value, what to emit when a value is unknown, and whether any prose is allowed around the payload. That last one is what usually breaks an integration: a valid object wrapped in a sentence of explanation is not parseable.
Is JSON the only shape covered?
No. The same path covers markdown, tables, lists and code — anything where the format is the requirement rather than a preference. The prompt declares the target format outright instead of leaving the model to infer it from the request.
Do they include an example of the output?
The structured path asks for one worked instance of the schema alongside the field list, because a list of fields with no instance beside it is where models start improvising key names.