How to tell if an image is Grok (Aurora)
Grok's image generation (xAI's Aurora model, and earlier a Flux integration) is known for photorealism and permissive content, and it adds no provenance metadata. Images saved from Grok usually carry a small, faint GROK mark in the bottom-right corner, but that is easy to crop, and once it is gone detection relies on the visual model plus the tell-tale absence of any camera fingerprint.
Visual tells
Photoreal but “metadata-naked”
Images look like photos yet carry no lens, ISO or device data — real camera shots usually do.
Flux-like fingerprints
Early Grok output shares Flux's characteristics; the same too-perfect micro-texture applies.
Compositing seams
Text-to-image edits of real photos can leave lighting or edge mismatches around inserted subjects.
The Grok watermark
Look at the bottom-right corner first. Images saved from Grok usually carry a faint, semi-transparent mark there: the word GROK next to a small square logo with a slash through it. On a busy background it is easy to miss, so zoom in on the corner.
The mark is part of the pixels, not the file’s metadata. A crop, a screenshot of part of the image or a quick paint-over removes it, and anyone can paste it onto a real photo. So a missing mark does not clear an image, and a mark alone does not prove it came from Grok.
A four-pointed sparkle in the corner is Google’s mark, not Grok’s: see how to tell if an image is Nano Banana. When there is no mark at all, upload the image below for the pixel check.
Metadata & provenance
| Signal | Present? | What it means |
|---|---|---|
| Content Credentials (C2PA) | No | No C2PA Content Credentials. |
| Visible watermark | Yes | A faint GROK wordmark and logo in the bottom-right corner. It sits in the pixels, so a crop or a paint-over removes it. |
| Invisible watermark | No | No documented invisible watermark. |
| Camera / EXIF fingerprint | No | No camera metadata; nothing to verify from provenance. |
How WeDetect flags Grok (Aurora) images
With no C2PA, our metadata checker finds no provenance, and the visible mark is often cropped before an image is shared, so the pixel model carries detection, supported by the missing-camera-metadata signal.
Check an image now
Drop the image in below — free, no sign-up, deleted after analysis.
Drag & drop, paste, or click to upload
PNG · JPG · WEBP · HEIC — your image is analyzed privately and deleted
This is an estimate, not proof — the detector can be wrong, producing both false positives and false negatives. Don't rely on it as your only source; verify anything important. wedetect.ai is not liable for decisions made from these results. See our Terms.
Frequently asked questions
Does Grok label its images?
Only visibly. Images saved from Grok usually carry a small GROK mark in the bottom-right corner, but no C2PA Content Credentials or documented invisible watermark, so once the mark is cropped you can't confirm origin from the file.
What is the small icon in the corner of Grok images?
It is the Grok logo: a small square with a slash through it, next to the word GROK, both semi-transparent. A four-pointed sparkle in the corner is a different mark, the one Google Gemini adds.
Is a Grok image a real photo if it has no EXIF?
No — missing EXIF is common to many AI generators and also to screenshots. Use the visual detector to judge.
Other generators: Midjourney · DALL·E 3 · GPT-4o Images · Stable Diffusion · Flux · all generators.