An AI image detection API for your product

What the API returns

A single POST /detect call with an image file returns a structured JSON response — no separate polling step, no webhook required for the basic flow:

{
  "scan_id": "scan_...",
  "verdict": "ai" | "real" | "uncertain",
  "confidence_level": "high" | "medium" | "low",
  "recommended_action": "allow" | "review",
  "evidence_summary": "...",
  "report_url": "/scans/.../report"
}

Common integration patterns

• Upload moderation — screen user-submitted images at upload time and route "uncertain"/AI results to a human review queue instead of auto-blocking.
• Marketplace trust & safety — check listing photos for AI-generated or manipulated images before publishing.
• Newsroom workflows — verify reader-submitted or social-sourced photos before publication, with a report link for the editorial record.
• Bulk/batch review — run existing image libraries through the API to flag likely AI content for audit.

Why recommended_action matters

The API doesn't just return a raw score — it returns a recommended_action of allow or review, so you can wire moderation logic without having to build your own probability-to-decision thresholds from scratch. High-confidence results are safe to automate; uncertain ones are flagged for a human, which keeps false positives from silently damaging real users.

There's a video endpoint too — /detect/video takes the same auth and returns the same response shape, plus a per-frame probability breakdown. See the AI video detector page for details.