JSON output AI API

If you want an AI API to return JSON output, the common approach is:

1. Ask for JSON in the prompt
2. Use a response format / schema option if the API supports it
3. Validate the returned JSON in your code

Example prompt

Return your answer as valid JSON with these keys:
- title (string)
- summary (string)
- tags (array of strings)

Example expected output

{
  "title": "Example",
  "summary": "A short summary.",
  "tags": ["ai", "api", "json"]
}

If you’re using an AI API with structured output support

Many modern APIs let you specify a JSON schema or “structured output” mode, for example:

• response_format: json
• response_format: json_schema
• function/tool calling
• strict structured responses

Best practices

• Make the schema explicit
• Tell the model to return only JSON
• Avoid markdown fences unless you really need them
• Validate and retry if parsing fails

Example Python check

import json

response_text = '{"title":"Example","summary":"A short summary.","tags":["ai","api","json"]}'
data = json.loads(response_text)
print(data["title"])

If you want, I can show you how to do this with a specific API like:

• OpenAI
• Anthropic
• Gemini
• Azure OpenAI
• a custom REST API