Catch fake listings before a buyer ever sees it.
Anyone can fake a listing photo or a seller ID in seconds. Witchunt suspends them immediately at upload in under 5 ms. Built for marketplaces where anyone can list.
Save money, time, and reputation.
Whether you're checking who's signing up, what they're listing, or what they upload during a dispute, we run the check at the upload point and pass the verdict to you.
New seller signup
Onboarding trigger
ID + selfie at upload
Captured client-side
Witchunt inspects asset
Verdict in 4 ms
Trust tier assigned
Seller risk model
Authentic
Approve seller
Continue onboarding
Synthetic / spoof
Suspend at signup
Before first listing
Eliminate fake sellers from your platform.
ID and selfie at signup. Re-checks when a seller gains a higher trust tier.
Works with the tools you already have.
One API call. We work alongside your seller verification, image storage, and trust-and-safety stack. Nothing to rip out, nothing to rebuild.
IDENTITY VERIFICATION
Persona
Veriff
Sumsub
FRAUD & TRUST
Sift
Sardine
SEON
MEDIA STORAGE
Cloudinary
AWS S3
imgix
The world's fastest and most accurate AI image and video detection.
New versions of AI don't scare us, and they shouldn't scare you either. We catch fake images and videos from AI tools we have never trained our models on. Never worry about the next version of AI again.
ProGAN
Acc
99.2%
First generation of AI images0 training samples
GenImage
F1
97.3%
Second (current) generation of AI imagesBest published score
DFDC
AUC
96.2%
Deepfake videoBest published score
DVF
AUC
94.7%
Mixed video forgeriesBest published score
See what we catch in your own listings.
Send us up to 200 of your listing photos or seller selfies. We send back a report in 24 hours. No signup, no commitment. Or request a demo to try it yourself.
Common questions.
Still curious? Email ahmed@witchunt.com.
AI-generated and edited images and video. Fully generated or partially altered photos and videos, plus fake IDs and documents.
Under 5 ms per image.
A verdict (real, tampered, or generated), a confidence score, the regions of the image that drove the call, and a short reason your T&S team can read. You can route on any of those.