DALL-E Image Detector: How to Identify OpenAI-Generated Images
DALL-E images from OpenAI can be identified through C2PA content credentials embedded in newer outputs, specific output dimension patterns, and the standard absence of camera EXIF metadata common to all AI generators.
DALL-E, developed by OpenAI and integrated into ChatGPT, is one of the most widely deployed AI image generators. DALL-E 3 (released 2023) and subsequent versions have introduced C2PA content credential support, which can make identification more straightforward, but only when credentials have not been stripped. Older DALL-E outputs and credentials-stripped images still require forensic metadata and dimension analysis.
C2PA content credentials in DALL-E images
OpenAI has committed to embedding C2PA (Coalition for Content Provenance and Authenticity) credentials into DALL-E outputs. C2PA is an open standard that embeds cryptographically signed provenance data into image files. A DALL-E image with intact C2PA credentials can be verified as AI-generated using any C2PA-aware tool. However, C2PA data is stored in the image's metadata and is removed when the image is saved, cropped, screenshot, or converted to a different format. Most DALL-E images found in the wild have had C2PA credentials stripped.
DALL-E output dimensions
DALL-E 3 outputs images at fixed resolutions: 1024x1024 (square), 1792x1024 (landscape), and 1024x1792 (portrait). DALL-E 2 outputs were 256x256, 512x512, or 1024x1024. All resolutions are multiples of 64. A JPEG or PNG at exactly 1792x1024 with no camera EXIF is almost certainly DALL-E 3 output.
Metadata patterns in DALL-E images
DALL-E images typically contain no camera EXIF fields. When saved through the ChatGPT interface, images may have a creation timestamp but no camera data. Some DALL-E outputs exported through the API or third-party integrations contain software metadata referencing OpenAI. When C2PA credentials are present, they include a 'c2pa.created' action with an AI generator reference.
Checking DALL-E images with the free detector
The ScanErase detector checks for missing EXIF, DALL-E standard dimensions (1024x1024, 1792x1024, 1024x1792), and sensor noise absence. An unmodified DALL-E 3 image at standard resolution with no EXIF will score 75 or higher, in the 'probable AI' or 'high confidence AI' tier.
Check any image with the free detector
Upload a photo to analyze its metadata, sensor noise, and pixel patterns. No account required. Your image never leaves your device.
Use free detectorFrequently asked questions
Does DALL-E embed watermarks?
DALL-E 3 embeds C2PA content credentials (cryptographic provenance data), not visible watermarks. These credentials are stripped when the image is screenshot, cropped, or re-saved in most software.
Can ChatGPT generate deepfakes?
OpenAI's usage policies prohibit generating realistic images of real individuals. However, policy violations occur, and other tools built on DALL-E models may not enforce the same restrictions.
Is a DALL-E deepfake covered by the TAKE IT DOWN Act?
Yes. The TAKE IT DOWN Act covers AI-generated intimate imagery regardless of the generation platform. DALL-E deepfakes have identical legal status to deepfakes created with any other tool.
Found a deepfake? Find every copy.
Detection is step one. ScanErase's biometric scan finds every copy across 200+ platforms, including social media, for $15, covering every platform simultaneously.
Start your scan