ScamCheck combines deterministic fraud detectors with AI vision and semantic matching to give you an explainable risk assessment in seconds. Here is the pipeline:
For text we parse the message directly; for screenshots we run OCR (with AI vision fallback) to extract the text — including Hindi and Hinglish.
We pull out links, UPI IDs, phone numbers, amounts, and QR/payment references — the things scams actually use.
Domains and emails are checked for typosquatting, homoglyphs, deceptive subdomains, and look-alikes of real brands; trusted/official entities are recognised.
Pattern detectors flag urgency, OTP-sharing requests, fake refunds/QR, KYC phishing, reward/job bait, and more — multilingual.
The extracted text is compared against a corpus of known scam campaigns for similar patterns.
Signals are combined and calibrated (anti over-confidence) into a risk score, a clear verdict, the evidence, and what to do next.
Privacy-first: images are optimized on your device and processed in-request, not stored. ScamCheck gives automated guidance, not legal/financial advice. Built by A Square Solutions.