Remove DALL-E Metadata from Images
Every image generated by DALL-E or ChatGPT contains hidden metadata that identifies it as AI-generated. This metadata triggers "Made with AI" labels on Instagram, Facebook, and LinkedIn. StripShot removes every trace at the binary level, with zero quality loss.
What DALL-E embeds in your images
OpenAI uses multiple layers of metadata to tag DALL-E output. Each layer serves a different purpose, and platforms check different layers. To fully remove AI detection markers, you need to strip all of them.
EXIF Software field
DALL-E writes openai and dall-e into the EXIF Software tag. This is stored in the APP1 marker of JPEG files and the EXIF chunk of PNG files. It is the simplest metadata to detect and the first thing platforms check.
Any EXIF viewer (ExifTool, Jeffrey's EXIF Viewer, even macOS Preview) can read this field. Social platforms scan it automatically on upload.
XMP metadata
OpenAI embeds XMP (Extensible Metadata Platform) data containing creator tool references, generation timestamps, and model identifiers. XMP is an XML-based format stored in a separate APP1 marker (distinct from EXIF).
The XMP data includes xmp:CreatorTool and dc:creator fields that reference OpenAI. Some AI detection tools specifically scan XMP for these fields, even when EXIF has been stripped.
C2PA Content Credentials (JUMBF)
This is the most important layer. DALL-E embeds C2PA (Coalition for Content Provenance and Authenticity) data using the JUMBF format (JPEG Universal Metadata Box Format). In JPEG files, this data lives in APP11 markers. In PNG files, it uses caBX, caMs, and caSt chunks.
The C2PA manifest contains a cryptographically signed declaration that the image was created by OpenAI's DALL-E model. This is exactly what Instagram and Facebook read to trigger the "Made with AI" label.
C2PA was developed by Adobe, Microsoft, Intel, and the BBC. It is the industry standard for content provenance tracking. Meta, Google, and TikTok have all committed to reading C2PA data.
OpenAI digital signatures
The C2PA manifest includes a digital signature chain that traces back to OpenAI's certificate authority. This signature proves the image came from DALL-E, not just any tool. StripShot removes the entire JUMBF box, including all signature data.
DALL-E vs ChatGPT image generation
Images generated through ChatGPT use the same DALL-E model and embed identical metadata. Whether you use the DALL-E API directly, ChatGPT's image generation feature, or the ChatGPT mobile app, the same EXIF, XMP, and C2PA data is present.
ChatGPT's image editing features (inpainting, extending, style changes) also add C2PA credentials to the output. This means even a real photograph edited through ChatGPT will receive AI metadata and could trigger platform labels.
The April 2025 ChatGPT image generation update (GPT-4o native image output) embeds the same C2PA Content Credentials as earlier DALL-E 3 output. The metadata format has not changed.
How StripShot removes DALL-E metadata
- Step 1: Binary scan identifies all APP1 markers containing EXIF data with "openai" and "dall-e" Software tags
- Step 2: XMP data in APP1 markers (separate from EXIF) is located and stripped entirely
- Step 3: C2PA JUMBF boxes in APP11 markers are identified by their JUMBF signature bytes and removed
- Step 4: For PNG files, caBX, caMs, and caSt chunks containing C2PA data are removed
- Step 5: Optional anti-fingerprint mode modifies pixel hash to defeat perceptual matching on social platforms
All steps happen in your browser. No server upload. Zero quality loss. Original format preserved.
Why screenshotting DALL-E images does not work
Taking a screenshot removes DALL-E metadata but creates new problems. Your device model, GPS location, and timestamp get embedded by your operating system. Image quality drops from compression. Resolution is limited to your screen size.
Screenshots also produce larger file sizes at lower quality compared to the original DALL-E output. A 1024x1024 DALL-E image is typically 200-400KB. A screenshot of the same image can be 2-5MB while looking worse.
StripShot preserves the full DALL-E output resolution, format, and quality. The only difference between input and output is the absence of metadata.
Platform detection by platform
| Platform | Detection method | Result |
|---|---|---|
| C2PA Content Credentials | "Made with AI" label added to post | |
| C2PA Content Credentials | "Made with AI" label added to post | |
| C2PA + EXIF scan | AI disclosure label on image posts | |
| Metadata + AI classifiers | Search suppression of AI-flagged pins | |
| Stock sites | EXIF + XMP scan | Submission rejected outright |
Related guides
Remove Midjourney Metadata
Strip Midjourney generation parameters and prompt data.
Remove Stable Diffusion Metadata
Strip SD parameters, prompts, seeds, and model hashes.
Remove Adobe Firefly Metadata
Strip Firefly C2PA Content Credentials and CAI data.
Remove All AI Metadata
Complete guide covering every major AI image generator.
Bypass AI Detection
Platform-by-platform guide for Instagram, Pinterest, Facebook.
Free. No sign-up required. Files never leave your device.