WordWatch API

Integrate craft-based AI detection into your platform.

Authentication

All API requests require a Bearer token. Get your API key from Settings.

curl -X POST https://wordwatch.ai/api/v1/scan \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Your text here...", "layer": 1}'

POST /api/v1/scan

Request Body

Parameter	Type	Description
`text`required	string	The text to analyze. Min 50 chars, max 100K chars.
`layer`	integer	1 = algorithmic only (default, free). 2 = deep analysis (Pro+).

Response

{
  "report": {
    "cas": 42,
    "verdict": "likely_ai",
    "confidence": 0.78,
    "categories": [
      {
        "name": "Rhythm & Structure",
        "score": 35,
        "label": "Concerning",
        "summary": "...",
        "signalCount": 3
      }
    ],
    "highlights": [
      {
        "id": "rhythm-flat-0",
        "text": "quoted passage...",
        "startOffset": 245,
        "endOffset": 512,
        "severity": "high",
        "signalIds": ["T3"],
        "evidence": "Found a run of 8 consecutive...",
        "source": "algorithmic"
      }
    ],
    "summary": "Strong AI-Generation Indicators...",
    "editorialNarrative": null,
    "meta": {
      "totalWords": 1243,
      "totalSentences": 67,
      "totalParagraphs": 12,
      "scanTimeMs": 23,
      "layer": 1,
      "analysisTokens": 0
    }
  }
}

GET /api/v1/scan/:id

Retrieve a previously saved scan result by ID. Only available for authenticated scans (Pro+).

curl https://wordwatch.ai/api/v1/scan/abc123 \
  -H "Authorization: Bearer YOUR_API_KEY"

Rate Limits

Parameter	Type	Description
`Free`		3 scans/day, Layer 1 only
`Pro`		100 API calls/month, Layer 1 + 2
`Enterprise`		Unlimited, custom rate limits

Verdict Values

Parameter	Type	Description
`human`	CAS 85-100	Authentic human writing
`likely_human`	CAS 70-84	Predominantly human writing
`mixed`	CAS 45-69	Mixed or indeterminate signals
`likely_ai`	CAS 25-44	Likely AI-generated
`ai`	CAS 0-24	Strong AI-generation indicators

Signal Reference

Each finding references one or more signal IDs from our 24-pattern forensics system.

Structural (S-series)

S1 — Frontloaded Qualification

S2 — Compulsive Comprehensiveness

S3 — Symmetrical Structure

S4 — Predictable Arc

S5 — Uniform Paragraph Length

S6 — Emotional Equilibrium

S7 — Absence of Personal Specificity

S8 — Philosophical Closer

S9 — Greeting/Validation Opener

S10 — Absence of Productive Imperfection

S11 — Uniform Sentence Openings

Telemetry (T-series)

T1 — Lexical Density

T2 — Transitional Scaffold

T3 — Rhythm Flattening

T4 — Abstract Noun Saturation

T5 — Hedge Cluster

T6 — Qualification Stacking

T7 — Rule-of-Three Compulsion

T8 — Tense Consistency Override

T9 — Pronoun Avoidance

T10 — Metaphor Recycling

T11 — Distributed Negation

T12 — Safety Disclaimers

T13 — Em Dash Dependency