Why does Claude prefer XML tags in prompts?

Anthropic's training data specifically highlights XML tags (like ) as a way to separate instructions from data, leading to higher reliability.

Structured Output (JSON & XML)

Structured Data: Turning Conversations into Code

Claude is the industry leader for Structured Data Extraction. Because of its meticulous adherence to instructions, it is the best model for turning "Messy Text" into "Clean JSON" for your software applications.

The Power of JSON Output

When you ask for a summary, you get a story. When you ask for JSON, you get Programmable Logic.

💻 Example: Automated Dashboard Data

Analyze this user bio and return ONLY valid JSON.

Schema:
{
  "name": "string",
  "industry": "string",
  "years_experience": number,
  "top_skills": []
}

Bio: [Paste Bio]

The Claude Special: XML Tags

One secret technique for Claude is using XML Tags within your prompt. This helps the AI understand where one part of the prompt ends and the next begins, which is critical for long-context tasks.

<context>
This is a transcript of a sales call between John and Sarah.
</context>

<instruction>
Identify the customer's main objection and return it in JSON format.
</instruction>

Top Use Cases for Structured Output

Automation: Passing data between AI and tools like Zapier or Make.com.
API Integration: Building apps where the AI acts as a data processor.
Database Cleanup: Parsing old, unstructured notes into a clean database table.

Structured Output (JSON & XML).