Prompt Engineering

JSON Mode Prompt — With and Without Native Support

Native JSON modes guarantee syntax, not your schema. The prompt contract that covers field names, types, and null discipline — whether or not the API has JSON mode.

Open in JSON Output Prompt Builder

Overview

OpenAI, Claude, and Gemini all offer some form of native structured output — and teams keep getting burned by assuming that solves the problem. Native modes guarantee the response parses; they don't guarantee it contains your field names, your types, or null instead of invented values. The prompt contract still does that work. This resource loads a support-ticket contract that works identically with JSON mode on (where it constrains content) and off (where it also forces syntax) — one prompt, both worlds.

How to use this resource

Generate the loaded contract

Note what native JSON mode can't do here: the severity value space, the null rule, the no-extra-fields rule.
Pair with your API settings

JSON mode on: keep the schema and validation sections, the syntax rules become redundant insurance. JSON mode off: everything earns its place.
Constrain values in descriptions

The severity field shows the pattern — enumerate legal values in the description and the model treats it as validation.

Why This Works

Separating syntax (the API's job) from schema discipline (the prompt's job) keeps both reliable
Value-space constraints in field descriptions are the cheapest enum enforcement available
One contract across providers means switching models is a config change, not a rewrite

Best for

API developers using response_format or structured output features
Teams running the same prompt on multiple model providers
Anyone who shipped JSON mode and still got wrong field names

Not for

Pure chat use where a human reads the output — the contract is overhead there
Validating responses after the fact — that's the AI Output Validator

Use cases

Writing one prompt that behaves across APIs with and without JSON mode
Constraining value spaces (severity: critical/high/medium/low) that native modes don't enforce
Migrating between model providers without rewriting the output handling

FAQ

If OpenAI or Claude already has a JSON mode, why still add this prompt contract?

Because native modes guarantee the response parses — not that it holds your field names, types, or null instead of invented values. This support-ticket contract does that: it names the schema, constrains severity to "critical, high, medium, or low," and requires reply_needed as a real boolean. With JSON mode on the syntax rules become redundant insurance; with it off they also force valid JSON. One contract, both worlds.

How do I get the model to respect an allowed set of values like severity levels?

Enumerate the legal values inside the field's line — the pattern the severity field shows: "critical, high, medium, or low" listed in the SCHEMA, which the model treats as validation. The whyThisWorks calls this "the cheapest enum enforcement available," since native JSON modes typically don't constrain value spaces. The prompt encodes it; run the output through a validator if your parser is strict.

Customize This Resource

Opens this setup in JSON Output Prompt Builder. Generate to get the full output contract — then adjust the schema, format, and strictness.

Open in JSON Output Prompt Builder

Inputs you need

Have these details ready before you copy or open the template.

The task the JSON output needs to support
The required fields, their types, and allowed values
Any validation rules or schema constraints
Example input data the model will convert, if you have it
How strict the downstream parser is (whether extra fields or nulls break it)

Prompt Template

Copy it as-is, or use Open in JSON Output Prompt Builder to load it pre-filled and customize it with your own context.

TASK
Convert the customer report in the input into a structured support ticket for our ticketing API.

OUTPUT FORMAT
Return your entire response as a single valid JSON object matching the schema below.

SCHEMA
Each field, its type, and its meaning:
- title (string, required): Short ticket title
- description (string, required): Problem description in the customer's words
- severity (string, required): critical, high, medium, or low
- product_area (string, optional): Affected product area
- reply_needed (boolean, required): Whether the customer expects a reply

EXAMPLE OF A VALID RESPONSE
{
  "title": "Concise descriptive title",
  "description": "Short free-text description",
  "severity": "high",
  "product_area": "standard",
  "reply_needed": true
}

VALIDATION RULES
- Every required field must be present.
- Optional fields with no value: set them to null — do not omit the key.
- Field types must match the schema: numbers without quotes, booleans as true/false.

OUTPUT RULES
- Output only valid JSON — nothing else.
- Do not include explanations.
- Do not include commentary.
- Do not wrap the output in markdown.
- Do not use code fences.
- Return exactly one JSON object.
- Follow the schema exactly — every required field present, no extra fields.
- Use null for unknown values — never invent data.
- No trailing commas.

More resources from JSON Output Prompt Builder

Resource

Force JSON Output from AI

Stop getting 'Sure, here is the JSON…' — the output-contract pattern that forces models to return only parseable JSON: schema, example, and a strict rule block.

Prompt Engineering

Resource

CSV Output Prompts

When the destination is a spreadsheet, CSV is the contract: one header row, one data row, quoting rules, and an honest answer about nested data.

Engineering

Resource

JSON Output for Zapier, Make, and n8n

Automation platforms fail silently on malformed model output. The strict contract that keeps AI steps feeding clean JSON into the rest of the flow.

Operations

Resources that pair well

Resource

Extract Data From Text with AI

Free text in, named fields out. The extraction prompt pattern that turns any unstructured text into consistent, parseable records.

Prompt Engineering

Resource

Fix Invalid JSON from AI

The JSON won't parse and you can't see why. Deterministic cause-sniffing — trailing commas, single quotes, unclosed brackets — and the repair prompt that fixes it.

Engineering

Resource

Fix a Vague Prompt

'Make it good', 'be detailed', 'keep it interesting' — vague prompts get vague output. The fix is mechanical: replace every fuzzy word with a checkable instruction.

Prompt Engineering

Related tools

Tool

JSON Output Prompt Builder

Build prompts that return structured data — JSON first, with YAML, XML, and CSV modes — parseable every time.

Structured Output

Tip: Save time by exploring related resources and tools that integrate with this resource.