Does your AI image generator leave metadata or Content Credentials?
Every AI image tool handles metadata differently. Some attach a signed C2PA Content Credentials record by default; some leave a plain-text generation record; some leave nothing at all. This does not by itself prove an image is AI-generated or genuine, inspect the actual file with the tools linked from each page.
Midjourney
What a Midjourney download from the web app or Discord does and does not carry in EXIF or C2PA Content Credentials.
Read →AI image generatorsChatGPT / DALL·E
OpenAI has documented adding C2PA Content Credentials to DALL·E 3 and ChatGPT image output since 2024.
Read →AI image generatorsAdobe Firefly
Adobe documents attaching Content Credentials (C2PA) to Firefly-generated images by default.
Read →AI image generatorsStable Diffusion
Popular open-source Stable Diffusion interfaces write prompt and sampler details into plain PNG text chunks by default.
Read →AI image generatorsGoogle Gemini / Imagen
Google applies its SynthID invisible watermark to Gemini and Imagen output, separate from visible file metadata.
Read →AI image generatorsOpenAI Sora
OpenAI has described attaching C2PA metadata and a visible watermark to content shared through the Sora product.
Read →AI image generatorsGrok (xAI)
No public documentation from xAI describes a Content Credentials or metadata policy for Grok image generation.
Read →AI image generatorsLeonardo AI
No clear public documentation describes a Content Credentials or metadata policy for Leonardo AI image exports.
Read →AI image generatorsIdeogram
No clear public documentation describes a Content Credentials or metadata policy for Ideogram image exports.
Read →AI image generatorsCanva Magic Media
Canva participates in the Content Authenticity Initiative broadly, but its Magic Media AI image output has no clearly documented per-file policy.
Read →Machine-readable version: /data/ai-generators.json