Romance scams are a growing problem on social media, and AI tools now make it easier for scammers to create convincing fake identities. A new study on arXiv describes a chatbot built to detect such scams in real-time through the messaging app Line. The system goes beyond simple keyword spotting by classifying where a conversation falls in the typical progression of a romance scam.
The research, posted as arXiv:2610.20631, focuses on stage classification—identifying early grooming, trust-building, and eventual exploitation phases. By catching scams earlier in the conversation, the chatbot could alert users before they are asked for money or personal information. The authors position this as a practical defense against AI-enabled fraud on popular chat platforms.
Because the abstract is brief, details on the chatbot's accuracy, training data, and deployment are not yet available. The work is an early signal that real-time, stage-aware detection may become a useful layer of protection for messaging app users.