Virtual communication extends beyond national borders and facilitates swift, ongoing contact, but it also tests the integrity of online interactions and raises questions about ethical conduct in digital spaces. In particular, dating platforms and unsolicited messages have the potential to exploit vulnerable individuals with the aim of extracting financial resources. The internet serves as a milieu where anonymity, disguises, and fraudulent behavior are widespread, creating conditions in which deceptive practices can flourish. Within this environment, scammers deploy linguistically informed devices designed to draw targets into fraudulent scenarios, using language as a tool to influence decisions and perceptions. To investigate these dynamics, the study screened and categorized persuasive linguistic devices drawn from seven letter-sets, each containing 18 to 23 letters, and included communications with scammers. The analysis revealed consistent lexical patterns and macrostructural configurations that are commonly associated with e-romance scams. The identified devices frequently signaled malicious intent, carrying a real potential to harm victims by steering them toward financial loss. These patterns suggest that the linguistic design of scam communications is not random but structured to maximize impact within the encounter. Among the most prevalent devices are strategies rooted in persuasion, flattering language, and appeals to trust, which tap into core human drives such as bonding, the allure of greed, and aspects of altruism. These elements operate within a broader macrostructure whose functional stages were clarified through a structural reduction approach, enabling researchers to map the sequence and roles of components within the letter-sets. The study thus offers a clearer picture of how linguistic design supports exploitation in online dating communications and provides a foundation for developing detection, prevention, and educational approaches to mitigate harm in digital social contexts.