Marketplaces & e-commerce
Marketplaces and shops rely on photos uploaded by sellers, renters and customers. Image generators make it easy to show an apartment that does not exist, a car nobody owns or a broken item that arrived intact. Our API scores those photos for AI generation so your team can review the risky ones first.
99.5%
of AI images caught with only 1% of real photos flagged, across 19 current generators
96.2%
still caught after WhatsApp, Instagram and re-share compression
0.987
AUROC on AI video from 12 video models
Hidden split of our October 2026 benchmark, never used in training. Method and full results
Scammers post rentals, cars and second-hand goods with generated photos, take a deposit and disappear. The pictures look like ordinary phone shots, so buyers and moderators trust them.
A customer can ask an image model to add a crack, a stain or a dent to a product photo and file a return or refund claim. Support agents have little time and no easy way to question the picture.
Sellers use generated product shots that misrepresent what ships, and review photos can be generated to make a product look popular or to attack a competitor.
A probability that each photo was generated or synthetically altered. Run it when a listing is created or edited, before it goes live.
Score the damage photos a customer attaches to a claim. A high score sends the claim to an agent with a reason to look closer before money goes out.
Photos in reviews are scored as images. Video reviews and walkthrough clips can be checked as whole clips, with frames sampled across the video and a count of how many looked AI-generated.
If a file carries a C2PA manifest, it is parsed and its signature verified. When a signer declares the image AI-generated, you get that as a verified fact instead of a model estimate.
Call the REST API when sellers publish a listing and when customers attach photos to a claim or review. Files up to 4MB can be uploaded directly; clips up to 50MB can be passed by URL.
You set the thresholds. Hold high-scoring listings for moderation, send flagged claims to a fraud or support agent, and let the rest through as usual.
Store the score on the listing, claim or review, so moderators and dispute teams can see what was flagged and act consistently on repeat offenders.
Score a photo attached to a refund claim
curl -X POST https://www.wasitaigenerated.com/api/v1/detect/image \
-H "Authorization: Bearer YOUR_API_KEY" \
-F "file=@claim-damage-photo.jpg"Images, single frames and whole clips. See the API docs for responses, limits and pricing.
The model looks for signs of AI generation and synthetic alteration. On our October 2026 benchmark it caught 99.5% of AI images while flagging 1% of real photos, across 19 current generators. Because every catalogue is different, we calibrate the threshold on a sample of your own images in a pilot.
Yes. The model is trained on images that went through messenger and social-media recompression, and still caught 96.2% of AI images after WhatsApp, Instagram and re-share compression on the same benchmark.
A detection score is a probability, not proof. Most teams use it to hold an item for review or ask for more evidence, such as another photo or a short video. Where a decision affects a customer or seller, a person should make it.
The REST API handles single requests and steady batch traffic. For high volumes we offer dedicated capacity, an SLA, self-hosting on your own GPUs and servers and a DPA. Uploaded files are deleted after analysis; the API keeps the result, not the file.
We run paid pilots on a sample of your images and clips, calibrate the threshold to your tolerance for false positives, and report detection and error rates back before you commit. Volume pricing and an SLA are available for production, and the models can run on your own GPUs and servers.
Talk to us about a pilot