Skip to content
LikelyAILikelyAI
Menu
Sign inGet API key

Documentation / Quickstart

Your first detection

Three steps. The result you get back carries a probability, a separate confidence in that probability, and the version of the detector that produced it.

  1. 1Create a test key

    Keys belong to an organization and are shown in full exactly once. Test keys are prefixed lk_test_ and never bill against your allowance.

    lk_test_01J8Q3ZK4F3ZK4F3ZK4F3ZK

  2. 2Send content

    One endpoint takes text, images, or audio. Pass exactly one of them. The Idempotency-Key header is required for anything billable, and repeating it returns the original detection instead of charging twice.

    curl https://api.likelyai.dev/v1/detect \
  -H "Authorization: Bearer $LIKELYAI_KEY" \
  -H "Idempotency-Key: req_01J8..." \
  -F "text=The keyboard has a reassuring weight."

  3. 3Read the result

    Fast operations return the completed detection directly. Longer ones, audio in particular, return status processing and you either poll GET /v1/detections/{id} or register a webhook.

    {
  "id": "det_01JQ8F3ZK4",
  "status": "completed",
  "modality": "image",
  "ai_probability": 0.87,
  "confidence": 0.74,
  "classification": "likely_ai",
  "signals": [
    { "name": "model_ensemble", "score": 0.91 },
    { "name": "content_credentials", "status": "not_found" }
  ],
  "model_version": "image-2026-08-1",
  "warnings": []
}

One thing to get right

ai_probability and confidence are different values. A result at 0.50 probability with high confidence means the content is genuinely ambiguous and well measured. The same 0.50 with low confidence means we could not measure that sample well. Treat them differently, and never collapse them into one number in your own interface.