OpenAI ChatGPT Watermark Checker

OpenAI ChatGPT Watermark Checker takes images or audio from ChatGPT and returns structured JSON on whether the file is AI generated. OpenAI uses Google DeepMind SynthID on ChatGPT images and audio. ChatGPT text is not watermarked, so this Actor does not accept pasted text or documents.

---

What is OpenAI ChatGPT Watermark Checker?

Upload a PNG / JPEG / WebP / GIF, or an MP3 / WAV / M4A / OGG clip. Each file becomes one dataset row: isAiGenerated, watermarkDetected, confidence, likely source, and billed token units.

No ChatGPT login. No browser.

---

What can this Actor do?

• Check ChatGPT images (PNG, JPEG, WebP, GIF)
• Check ChatGPT audio (MP3, WAV, M4A, AAC, OGG, FLAC)
• Return isAiGenerated plus SynthID / watermark fields in structured JSON
• Charge per file and per 1,000 input tokens so a long clip costs more than a small still
• Export JSON, CSV, or Excel, or pull results via the Apify API

It does not classify essays, PDFs, or other ChatGPT text. There is no SynthID-Text signal to read there.

---

Why check ChatGPT images and audio?

ChatGPT can emit stills and voice. A watermark check will not replace editorial judgment, but it gives you a machine-readable signal you can log next to the file.

Typical jobs:

1. Newsroom provenance — flag ChatGPT images before they ship
2. Brand safety — scan UGC photos and clips at moderate volume
3. Marketplace QA — catch generated product shots
4. Trust and safety — label synthetic voice notes
5. Legal extras — attach isAiGenerated to produced media
6. Dataset labeling — tag training images as AI vs camera
7. Publisher intake — check freelance illustrations
8. Education tools — show a confidence score, not a red card
9. Ad review — screen generated creatives
10. Internal audits — sample Slack/Drive media dumps
11. Policy logging — keep a synthetic-media field
12. Research studies — batch-label corpora of stills and audio

---

How to use OpenAI ChatGPT Watermark Checker

1. Open the Actor in Apify Console
2. Upload Files (images or audio)
3. Click Start
4. Download the dataset or read it via the Apify API

Input example

{
  "files": ["https://example.com/chatgpt-still.png"],
  "maxItems": 5
}

Field	Type	Default	Description
files	string[]	—	Image or audio URLs to check
maxItems	integer	5	Cap on files this run

---

Output

Each dataset item is one file. The headline field is isAiGenerated. model is the detector family (synthid).

{
  "id": "a1b2c3d4e5f67890",
  "sourceType": "image",
  "sourceName": "still.png",
  "sourceUrl": "https://api.apify.com/v2/key-value-stores/.../records/still.png",
  "imageUrl": "https://api.apify.com/v2/key-value-stores/.../records/still.png",
  "mediaKind": "image",
  "mediaType": "image/png",
  "fileSizeBytes": 184320,
  "isAiGenerated": true,
  "watermarkDetected": true,
  "watermarkType": "synthid",
  "matchesTargetProvider": true,
  "likelySource": "chatgpt",
  "likelihood": "high",
  "confidence": 0.82,
  "detectionMethod": "hybrid",
  "rationale": "OpenAI SynthID signal on a ChatGPT-style still.",
  "limitations": null,
  "inputTokens": 1290,
  "outputTokens": 160,
  "billedTokenUnits": 2,
  "model": "synthid",
  "targetProvider": "openai-chatgpt",
  "error": null,
  "checkedAt": "2026-08-28T12:00:00.000Z"
}

---

Pricing

Larger images and longer audio use more tokens. This Actor passes that cost through with a margin.

Event	When	Price (FREE)
text-checked	Once per successful file	$0.005
input-1k-tokens	Per 1,000 input tokens actually sent (rounded up)	$0.001
Actor start	Per run (× memory GB)	platform default

Failed downloads are not billed.

Examples (excluding start fee):

• Small JPEG (~1,000 tokens) → ≈ $0.006
• Larger still or short clip (~3,000 tokens) → ≈ $0.008

---

Files

• Images: PNG, JPEG, WebP, GIF
• Audio: MP3, WAV, M4A, AAC, OGG, Opus, FLAC
• Files over 15 MB are rejected
• PDF, Word, and plain text are rejected — ChatGPT text has no SynthID watermark

For Google SynthID-Text on writing, use Google Gemini Watermark Checker. For Claude-style copy, use Anthropic Claude Watermark Checker.

---

Is it legal to check AI-generated media?

This Actor only analyzes files you provide. It does not log into ChatGPT. Do not treat the score as legal proof of authorship. A watermark estimates whether ChatGPT media was involved; it does not identify a person.

---

Tips

• Prefer the original export, not a heavily recompressed screenshot
• Watch billedTokenUnits if a run costs more than you expected
• Start with maxItems 1–3 while you learn the output shape

---

Image Credit

Image credit: OpenAI / ChatGPT