Metadata can be removed or edited. This tool reports available metadata, rather than authenticating an image.
How AI Image Detection Works
Generative models from OpenAI, Google, and Midjourney use different signals to identify synthetic content. Here is how detection works in practice.
The Three Main Signals
- 1. 官方 C2PA 数字凭证 (Content Credentials)：Open standard supported by OpenAI, Adobe, and Microsoft. Embeds a provenance manifest that may describe camera capture, editing, or AI generation. A C2PA marker alone does not establish AI origin, and this tool does not verify digital signatures.
- 2. 嵌入的生成参数 (Prompt / Workflow)：PNG files from Stable Diffusion, ComfyUI, or Midjourney often preserve prompts, negative prompts, and seeds in text chunks.
- 3. 相机硬件 EXIF 参数：Real camera photos contain physical camera sensor parameters. Synthetic images typically lack hardware signatures.
Why Does Editing or Sharing Strip Detection Signals?
Social media platforms compress images and automatically strip metadata and C2PA credentials to save bandwidth. Similarly, screenshotting or editing an image creates a new file without original provenance signals.