Get started

Prerequisites#

• A Defacta account and API access
• A piece of AI-generated text you want to check
• A backend, agent, or browser where the result will be used

1. Choose how you will call Defacta#

Pick the surface that fits your workflow: the REST API for backends and pipelines, the MCP server for agents, or the browser extension for in-place checks. You can use more than one.

2. Submit text for verification#

Send the content you want checked. Defacta breaks it into individual factual claims that can be evaluated independently.

3. Read the structured verdict#

Defacta returns a JSON payload summarizing which claims are supported, which are contradicted or unverifiable, and what was flagged for bias or manipulation.

4. Act on the result#

Gate an AI response on the verdict, log it for audit, or surface it inline in your UI - all from structured fields, no free-text parsing.

Expected result#

You have turned an unverified block of AI text into a machine-readable trust signal your product can branch on.

Next: choose your integration.