Tell real photos from generated ones
Upload a photo, illustration or screenshot and see how likely it is that an image model made it, with a heatmap of the regions that gave it away.
Free to try with an account. JPG, PNG and WebP files.
How it works
From upload to an answer you can explain, in three steps.
Step 1: Upload an image
Drop in a JPG, PNG or WebP file, or paste an image link. Screenshots from social feeds work too.
Step 2: We look for generator traces
The model checks pixel-level noise, lighting and texture patterns that image generators leave behind, and reads any metadata or content credentials attached to the file.
Step 3: Read the score and heatmap
You get a likelihood score plus a heatmap of the regions that drove it, so you can judge the result with your own eyes.
Evidence, not just a percentage
Every scan shows where the signal came from, so a result can be checked, questioned and explained.
Region heatmap
See which parts of the image look generated, from skies and skin to backgrounds that repeat.
Metadata and credentials
We read camera data and content credentials when they’re present, and tell you when they’ve been stripped.
Partial edits
Catches real photos where one area was regenerated or an object was added with an AI fill tool.
Major generators covered
Trained on output from widely used diffusion and image models, and retrained as new ones are released.
Reports to share
Download a PDF with the score, heatmap and file details for your records or an editor’s review.
Private uploads
Images are deleted after the scan unless you save the report. We never train on your uploads.
Who uses it
Anyone who has to decide whether an image is what it claims to be.
Newsrooms and fact-checkers
Verify before you publish
Check reader-submitted photos and viral images before they run, and keep a report on file with the story.
Art and design teachers
Portfolios you can trust
Check a piece when it doesn’t match a student’s earlier work, then talk it through with the evidence in hand.
Marketplaces and trust teams
Profiles and listings
Screen profile photos, product shots and review images for generated content at scale through the API.
Image detector questions
Something else on your mind? Visit the help center or ask our team.
Sometimes it can point to a family of generators, but the main answer is how likely the image is to be AI-made. We don’t present a guess about one specific tool as fact.
Yes, though heavy compression, resizing and filters remove some of the traces we look for. Small or heavily edited images get a lower confidence level, and the result says so.
Partial edits, such as a replaced sky or an object added with generative fill, usually show up on the heatmap as one bright region rather than a high score across the whole image.
No. Treat it as a strong reason to look closer: find the original source, check other photos from the same person or event, and compare details like shadows and text.
They’re analyzed and then deleted, unless you save the report to your account. We never use your images to train our models.
Know what you’re looking at before you share it
Create a free account and check your first images in seconds.