Safe Sharing: When to Strip Midjourney Metadata, How to Keep Internal Provenance

Midjourney has solved the metadata problem. As of March 2026, every image you download from midjourney.com—single or batch—arrives with your full prompt, generation parameters, Job ID, your Midjourney username, and an IPTC Digital Source Type tag, all embedded in the PNG metadata. This is genuinely useful for archiving: your files are self-documenting, searchable, and traceable back to the exact generation job.

But the same metadata that makes your archive useful makes every shared file a potential privacy leak. When you send an image to a client, post it on social media, or include it in a portfolio, anyone who inspects the file with ExifTool or an online EXIF viewer can read your complete creative process: the exact prompt, every parameter, your username, and the generation timestamp.

This article provides a practical framework for deciding when to strip metadata, when to preserve it, and how to do both without losing your internal record.

What's Actually in the Metadata

Before deciding what to strip, you need to know what is there. Every current Midjourney download embeds these fields:

• Description — A single text field containing the full prompt, all parameters (--ar, --stylize, --chaos, --v), and the Job ID. This is not structured into separate fields; it is one long string that must be parsed to extract individual values.
• Digital Image GUID — The Job ID as a UUID, enabling programmatic matching against your midjourney.com generation history.
• Author — Your Midjourney username. This identifies you personally in every file you share.
• Creation Time — The timestamp of the original generation.
• IPTC Digital Source Type — The value trainedAlgorithmicMedia, an industry-standard tag that marks the image as AI-generated.

Single downloads and batch ZIP downloads contain identical metadata. There is no difference in what is embedded between the two methods. If a file was downloaded from midjourney.com after late 2025, it contains all five fields.

The Privacy Problem: Three Real Risks

Embedded metadata is invisible unless you look for it. Most people never inspect EXIF data. But the people who do are often the ones who matter most.

Risk 1: Competitive advantage leaks. If you have spent weeks refining a prompt that produces a distinctive style, that prompt is embedded in every output. A competitor, a collaborator, or a curious follower can extract it in seconds. Your creative IP leaves with the image.

Risk 2: Client confidentiality breaches. Prompts often reference specific brands, product names, campaign briefs, or confidential project details. A prompt like “photorealistic product shot of [Client]'s unreleased headphone model, matte black, studio lighting” reveals the client relationship, the product, and the fact that AI was used—all in one metadata field.

Risk 3: Involuntary AI disclosure. The Author field (your MJ username) and the IPTC Digital Source Type tag make AI origin discoverable by anyone who inspects the file. Depending on context, this may be welcome transparency or an unwanted reveal.

When to Strip: The External-Facing Rule

The general principle is straightforward: strip prompt metadata from any file that leaves your organisation. The specific scenarios:

• Client delivery — Clients need the image, not your prompt. Remove the Description and Author fields before delivery. Keep IPTC Digital Source Type if AI disclosure is required (increasingly, it is).
• Portfolio and case studies — Published work should showcase results, not reveal the techniques that produced them. Strip prompts to protect your creative methodology.
• Social media — Some platforms strip EXIF on upload; others preserve it. Do not rely on the platform to protect you. Strip before uploading.
• Stock or marketplace submissions — Images uploaded to stock libraries or creative marketplaces should not carry your prompt text or username.
• Press or media assets — Journalists and researchers routinely inspect metadata. If you do not want your prompts in a news article, strip before sharing.

When to Keep: The Internal Archive Rule

Metadata is valuable. Stripping it indiscriminately is as problematic as leaving it everywhere. The rule: preserve full metadata in your internal archive.

• Internal archive and DAM — Your master copy should retain all metadata. This is your searchable record: find images by prompt text, filter by parameters, match to Job IDs.
• Compliance records — The IPTC Digital Source Type tag (trainedAlgorithmicMedia) is your evidence of AI origin. If regulations require AI disclosure, this is the field that proves it. Never strip this from compliance copies.
• Team collaboration — When sharing internally, prompt context helps team members understand creative decisions, iterate on existing work, and avoid duplicating effort.
• Audit trail — For client work, your internal records should show exactly what was generated, when, and with what parameters—even if the delivered version is stripped.

The Redact vs Preserve Matrix

* “Keep” for IPTC Digital Source Type in external deliveries means: leave the AI-generated tag intact when AI disclosure is required or expected. This is increasingly the case under EU AI Act guidelines (enforcement from August 2026) and client contracts that mandate AI transparency.

How to Strip Selectively with ExifTool

ExifTool is the standard command-line tool for metadata manipulation. Here are the specific commands for Midjourney privacy management:

Strip prompt and author, keep IPTC source type and creation time:

exiftool -Description= -Author= -overwrite_original image.png

This removes the prompt text and your username while preserving the IPTC Digital Source Type tag and Creation Time. The image retains its AI-generated disclosure without revealing your creative process.

Strip everything (nuclear option):

exiftool -all= -overwrite_original image.png

This removes all metadata. Use only when you need a completely clean file with no traceable information. Note that this also removes the IPTC Digital Source Type tag, which may cause compliance issues.

Batch strip for a delivery folder:

exiftool -Description= -Author= -overwrite_original /path/to/delivery/

This processes every file in the folder. Use this when preparing a client delivery package where you want consistent metadata handling across all assets.

The Unsolved Problem: Stripping Is Irreversible

Here is the fundamental tension: once you strip metadata from a file, it is gone. You cannot re-embed a prompt into an image after the fact (not accurately, anyway). This creates a strict requirement: you must have an internal system of record before you strip anything.

The workflow must be: download with full metadata, import into your archive or DAM, then export stripped copies for external use. If you strip first and archive second, you have lost the very data that makes your archive searchable and auditable.

This is not a hypothetical risk. Teams that strip metadata as part of their standard download workflow—or worse, rely on social media platforms to do it for them—end up with an internal library of images they cannot search by prompt, cannot trace to specific jobs, and cannot prove AI origin for compliance.

A post-export management system solves this by maintaining the full metadata internally and producing stripped exports on demand. The master copy never changes. Each export applies a privacy preset that controls exactly which fields are included. You get both: full internal provenance and safe external sharing.