Midjourney detector
Midjourney images are recognisable in two ways: what the file records about itself, and what the pixels look like statistically. This tool checks both, starting with metadata because it is the only part that can give a definite answer.
Midjourney output downloaded straight from Discord or the web app often retains identifying metadata, and the standard output dimensions are a further hint. Once an image has been screenshotted or passed through a social platform, those traces are gone and only the classifier remains.
Drag a file here, or
JPEG, PNG, WebP, AVIF, PDF or .txt · up to 10 MB · you can also paste a screenshot
By uploading you agree to our Terms and Privacy Policy. Your file, IP address, approximate location and device details are stored.
What this checks
- EXIF and XMP fields naming Midjourney or its version
- Midjourney's characteristic output dimensions and aspect ratios
- Absence of any camera fingerprint where a photograph would have one
- Two image classifiers trained to separate diffusion output from photography
Frequently asked questions
Does Midjourney add metadata to its images?
Often, yes — images taken directly from Discord or the web gallery commonly carry identifying fields. Screenshots and re-uploads do not, which is why the same image can give a confident answer in one form and an uncertain one in another.
Can it tell Midjourney apart from other generators?
Only when the metadata names it. The classifiers detect that an image is diffusion-generated, not which product produced it. If the metadata is gone, expect 'AI-generated' rather than 'Midjourney'.
Why does a Midjourney image sometimes score as uncertain?
Because recent Midjourney output is photorealistic enough to sit outside what the detection models were trained on, and if the metadata has been stripped there is nothing conclusive left to read.