Social media has become a breeding ground for scammers, and a new study reveals just how sophisticated their tactics have become. Researchers have developed an automated system called ScamChatBot, which mimics real users seeking help with account recovery on platforms like Facebook, Instagram, and X (formerly Twitter), as well as cryptocurrency wallets. What they discovered is alarming. By deploying this chatbot, researchers engaged with over 11,700 scammers, ultimately having in-depth conversations with 450 of them. The bot, powered by a large language model (LLM) like ChatGPT, was incredibly effective at luring scammers into revealing their modus operandi, including the payment methods they use to collect money from victims. The research uncovered that scammers often employ multiple social media profiles, sometimes as many as 71, to create a sense of legitimacy and urgency. They respond quickly to potential victims, often within two hours, and use various social engineering tactics, including impersonating official support staff, requesting personal information for “verification,” and inventing technical problems that require payment to fix. Disturbingly, the study also found evidence of scammers using extortion and threats if victims hesitate to pay. Furthermore, the analysis of the scammers' text revealed that nearly 30% are likely using AI-powered text generation tools to enhance their deception. The researchers shared their findings with platforms like X and PayPal, as well as the cryptocurrency abuse database Chainabuse. The information led to the identification and remediation of over 3 million scam accounts on X alone, highlighting the scale of the problem. This research demonstrates not only the prevalence of social media scams but also the potential of automated systems like ScamChatBot to combat them, paving the way for more sophisticated detection and prevention methods in the future.