# WasItAIGenerated

> AI content detection for text, images, documents, voice and video. Web app plus a REST API. Text detection runs on tropa-2, our own model; results come with a calibrated 0–100 score, a verdict band, per-sentence scores, a "possibly humanized" hint and a hidden-character warning.

Site: https://wasitaigenerated.com
API docs: https://wasitaigenerated.com/docs
OpenAPI: https://wasitaigenerated.com/openapi.json
Contact: patrick@wasitaigenerated.com · https://wasitaigenerated.com/contact

## When to use this service

Use it when you need to know whether a piece of text, an image or a document was produced by an AI model:

- Screening submissions (essays, applications, articles, reviews) before a human looks at them
- Verifying images and documents (C2PA content credentials, provenance signals, AI-generation likelihood)
- Building moderation, academic-integrity or fraud-review workflows that need a machine-readable score
- Checking whether AI text was post-processed with a humanizer tool

Do not use it as the sole basis for an accusation. Results are evidence, not proof: route `uncertain` verdicts to a human, and show flagged sentences to the person concerned.

## How an agent should call it

1. Get an API key: sign up at https://wasitaigenerated.com/sign-up, verify the email (1,000 free credits), copy the key from https://wasitaigenerated.com/dashboard. Keys start with `wai_`.
2. Send `Authorization: Bearer <key>`.
3. Text: `POST https://wasitaigenerated.com/api/v1/detect/text` with JSON `{"content": "<text>", "mode": "depth"}`. Returns `isAI`, `confidence`, `score` (0–100), `verdict` (`human` | `uncertain` | `likely_ai` | `ai`), `sentences[]`, optional `humanizedHint`, `inputAnomalies`, `subclassProbs`.
4. Image: `POST https://wasitaigenerated.com/api/v1/detect/image` as `multipart/form-data` with field `file` (PNG, JPG, GIF, WebP, ≤ 4 MB). Returns `isAI`, `confidence`, `verified`, `provenance`, `c2pa`.
5. Cost: text 1 credit per word, image 300 credits. `402` means the balance is empty; `403` means the plan has no API access.
6. Timeouts: allow 30 s. Responses are JSON; errors are JSON with `error`, `message` and `hint`.

Full request/response schemas with operationIds: https://wasitaigenerated.com/openapi.json

### Minimal example

```bash
curl -X POST https://wasitaigenerated.com/api/v1/detect/text \
  -H "Authorization: Bearer wai_..." \
  -H "Content-Type: application/json" \
  -d '{"content": "Paste at least ~50 words here.", "mode": "depth"}'
```

## Reading a result

- `score` is calibrated: 62 means roughly six in ten texts scoring there are AI-generated.
- Bands: `< 40` human · `40–69` uncertain · `70–89` likely AI · `≥ 90` AI-generated. `isAI` is `score > 50`.
- `humanizedHint: "possibly_humanized"` is only ever present, never `false`. Its absence means nothing.
- `inputAnomalies` lists hidden characters (homoglyphs, zero-width, exotic spaces) that were stripped before scoring — often a sign of deliberate evasion.
- `model` names the model that scored the text (`tropa-2`; `tropa-1` on request via `"model": "tropa-1"`).

## Pricing

- Free: 1,000 credits after email verification
- Starter Pack: $5 for 25,000 credits (one-time, never expires)
- Credit Package: $29 for 200,000 credits (one-time, never expires)
- Unlimited Website: $14.99/month, web app only, no API
- Unlimited API + Website: $299/month
- Enterprise: custom — patrick@wasitaigenerated.com

## Accuracy, honestly

Measured August 2026 on held-out data: 97% of raw output from current models (GPT-5.6, Claude Opus 5, Gemini 3.6 …) detected at a 0.5% false-positive rate on human text; humanized AI text up to 78%. Full benchmark with public data: https://wasitaigenerated.com/research/ai-detector-benchmark-gptzero-zerogpt-tropa-2-2026

## Key pages

- /docs — API reference
- /openapi.json — machine-readable API specification
- /ai-text-detection-api, /ai-image-detection-api, /chatgpt-detection-api, /deepfake-detection-api — API guides per use case
- /research — benchmarks and technical write-ups
- /research/tropa-2-detection-model — the current text model
- /education — for teachers and institutions
- /academic-ai-detector — AI detection for essays, research papers, and theses
- /thesis-check — thesis review with AI score
- /about, /contact, /privacy, /terms, /imprint

## Company

WasItAIGenerated.com, Stettiner Str. 41, 35410 Hungen, Germany. Founded by Patrick Gerard. VAT ID DE351122428.