Midjourney Metadata: What's Embedded and How to Remove It

Midjourney is the most widely used AI image generation platform in the world, with millions of creators using it daily to produce images for commercial projects, social media, marketing materials, and personal art. What many of these creators do not realize is that every Midjourney image — whether downloaded from the Discord bot or the Midjourney web interface — contains embedded metadata that identifies it as AI-generated.

This guide focuses specifically on Midjourney's metadata practices: what data is embedded in Midjourney outputs, how this data is used by detection systems, and the most efficient workflow for removing it. If you use Midjourney professionally, this information is essential.

What Metadata Does Midjourney Embed?

Midjourney's metadata practices have evolved across versions. In Midjourney V5 and V6, images downloaded from Discord contain EXIF data with the Software field set to 'Midjourney' (or a version-specific string like 'Midjourney V6.1'). The UserComment field may contain partial prompt information. The Artist field is sometimes populated with the Midjourney account username.

Images downloaded from the Midjourney web interface (midjourney.com) in 2025-2026 contain more comprehensive metadata. The XMP packet includes a dc:creator field with the account username, a dc:description field with the full prompt, and an xmp:CreatorTool field identifying Midjourney. The IPTC Application Record may include the prompt as a caption and the account username as a byline.

How Detection Tools Use Midjourney Metadata

AI content detection tools use Midjourney metadata in several ways. The most straightforward is keyword matching: scanning the Software, UserComment, and XMP:CreatorTool fields for known AI tool names. This is the detection method used by most stock photography platforms and basic content auditing tools. It is also the easiest to defeat with metadata removal.

More sophisticated detection involves cross-referencing the account username found in XMP:dc:creator with known Midjourney accounts. If a platform has a database of Midjourney usernames (which is possible given that Midjourney operates through Discord, where usernames are public), it can potentially identify images as Midjourney-generated even without the Software field. This is a more advanced detection method that requires removing the username fields specifically.

Beyond file metadata, Midjourney images also have characteristic pixel-level signatures. The V6 model in particular produces images with distinctive sharpness patterns, color saturation characteristics, and texture rendering that trained classifiers can identify. Addressing this requires pixel fingerprint modification in addition to metadata removal.

The Fastest Workflow for Cleaning Midjourney Images

For creators who regularly use Midjourney for commercial work, efficiency is as important as thoroughness. The following workflow balances both:

1. 1Download your Midjourney images in the highest available resolution (use the 'Upscale' options in Discord or the web interface).
2. 2Upload all images to BlankAI — you can process up to 20 images simultaneously. BlankAI removes all Midjourney metadata fields (EXIF, XMP, IPTC) and applies pixel fingerprint modification in a single step.
3. 3Download the cleaned images from BlankAI's results panel. Each image will have a new SHA-256 hash and no AI-identifying metadata.
4. 4If needed, open the cleaned images in your photo editor for any additional post-processing (color grading, cropping, retouching).
5. 5Use BlankAI's Image Diff tool to verify that the metadata has been removed and the hash has changed before submitting to clients or platforms.

Midjourney V6 vs. V5: Metadata Differences

Midjourney V6, released in late 2023 and now the default model, embeds more comprehensive metadata than V5. V6 images from the web interface consistently include the full prompt in XMP fields, while V5 images from Discord often only included partial prompt data. This means that V6 images require more thorough metadata removal — specifically, the XMP packet must be completely removed, not just the EXIF fields.

The Canvas re-encoding method used by BlankAI handles this automatically: because Canvas only processes pixel data, the resulting file contains no XMP, EXIF, or any other metadata container, regardless of which Midjourney version generated the original image.

Does Midjourney Use C2PA?

As of early 2026, Midjourney does not embed C2PA content credentials in its outputs. This is in contrast to DALL-E 3 and Adobe Firefly, which do embed C2PA by default. However, Midjourney has been a member of the Content Authenticity Initiative since 2023, and there is ongoing speculation that future Midjourney versions may add C2PA support. For now, Midjourney metadata removal does not require C2PA-specific handling — but using a tool that handles C2PA (like BlankAI) ensures you are covered if this changes.