AI Image Detector
Check a JPG, PNG or WebP for signs of AI generation. See the image classification, its confidence and a heatmap when available.
Check a photo
Read your detection result
- Classification
- Start with the category. A result can be likely AI-generated, likely real or uncertain. Edited, unknown and low-confidence results can stay uncertain.
- Confidence
- The percentage tells you how strongly the detector supports the returned category. It does not measure how much of the image was generated.
- Heatmap
- When a map is available, compare it with the original to see which regions the detector flagged. Highlighted areas are clues for review.
How to check an AI-generated image
Use the best available file
Prefer the original download over a screenshot. Cropping, resizing and repeated compression can remove or alter useful signals.
Run one image check
Upload a supported file and confirm the analysis. Free and paid checks use the same detector and interpretation thresholds.
Verify beyond the score
Look for the earliest source, creation records and Content Credentials. Treat the detector as one source of evidence, not a certificate.
What the result cannot prove
A result cannot identify the photographer, prove copyright ownership, verify the event shown or reliably name the generator that created the image. New models and editing workflows can also change detector performance.
Can an AI image detector identify Midjourney or DALL·E?
You can check images made with these tools, but the current detector does not provide verified generator attribution. A category result cannot prove which model or model version created an image.
Does the detector work on screenshots?
Screenshots can be uploaded as JPG, PNG or WebP, but recompression and reduced resolution may weaken the available evidence. Use the original file whenever possible.
Is a high score proof that an image is AI-generated?
No. A high confidence score supports the returned classification, but false positives and false negatives remain possible. Confirm the source and context before making a consequential claim.