Recruiting
Remote hiring means many candidates are only ever seen on a screen. Some of those faces are generated, and some recorded applications are deepfakes. Hiring platforms and HR teams use our API to score profile photos, recorded video and screen grabs for AI generation, then have a person look at anything flagged.
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
Fake candidate profiles use AI-generated headshots that look like a professional photo shoot. A reverse image search finds nothing, because the face was never photographed.
People apply under invented or borrowed identities to get access to a company, its systems or its payroll. In a fully remote process, the photo and the video may be the only evidence of who is applying.
One-way video interviews and recorded applications can be altered with face-swap and video generation tools. Recruiters watching dozens of clips are not in a position to spot it.
A probability that a candidate photo was generated or synthetically altered, from a model trained on current image generators.
Upload a recorded clip or pass its URL. Frames are sampled across the whole clip and scored by a dedicated video model, and the result reports how many frames were analysed and how many were flagged.
The API does not join or analyse live calls. After an interview, your team can check the recording, or a few screen grabs of the candidate, as single frames with the video model.
If a photo or clip still carries a C2PA manifest, it is parsed and its signature verified, and a valid AI declaration from the signer is reported as such.
Call the REST API when a candidate uploads a photo or submits a recorded application, or after an interview with the recording or screen grabs.
Choose your own thresholds. A high score should trigger a follow-up, such as an extra verification step or a second interview, and a recruiter should review the case.
Store the score and frame counts with the application, so the reason for a follow-up is clear to the hiring team and to the candidate if they ask.
Score a recorded video application
curl -X POST https://www.wasitaigenerated.com/api/v1/detect/video \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"url": "https://example.com/applications/candidate-intro.mp4"}'Images, single frames and whole clips. See the API docs for responses, limits and pricing.
No. The API checks recorded clips and still frames. It does not analyse live calls or streams. If you record interviews, you can send the recording afterwards, or take a few screen grabs during the call and check them as single frames.
A detection score must never be the sole reason to reject a candidate. It is a probability, not proof. Use it to decide when to ask for more verification, and let a person make the hiring decision based on everything you know.
On our October 2026 benchmark the video model reached an AUROC of 0.987 across 12 video models. It caught 60 of 63 AI clips and flagged 1 of 60 real clips. Results on your own recordings depend on their quality, so we recommend a pilot to set the threshold.
Uploaded files are deleted after analysis; the API keeps the result, not the file. A DPA, dedicated capacity and self-hosting on your own GPUs and servers are available for HR teams with strict data requirements.
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