Eight signal families
The model checks language and link structure for account theft, money requests, threats, urgency, impersonation, isolation, unexpected rewards and suspicious actions.
ShieldHer AI Scam Signal
Paste a suspicious message, link or contact detail.
Our original model looks for pressure, impersonation, payment demands, account theft, extortion and risky link structure. It explains every signal instead of quietly assigning a verdict.
Messages and links accepted togetherURLs never opened during analysisNo submission or message storage
Try an example
Explainable by design
A phone number, email or unfamiliar link does not create a scam verdict by itself. The score rises when meaningful warning signs appear together.
A screening result
The score measures detected warning signals, not the probability that a person is a criminal. New and carefully written scams may avoid detection.