Before the money disappears, there is a vocabulary. And it is remarkably consistent.
Romance scams are one of the few crimes where the entire attack is built from words. There is no malware, no stolen password, no forged document. Just a persona, a script, and enough patience to make both feel real. That makes this genuinely fertile ground for forensic linguistics, because the language itself carries the evidence.
According to the Federal Trade Commission, romance scams produced the highest reported losses of any imposter scam category it tracks, with a median loss per victim of $2,000. As a U.S. federal government publication, this figure is public record and free to cite without restriction. The FTC's broader Consumer Sentinel Network data shows romance scam losses climbing from roughly $200 million in 2019 to well over a billion dollars in more recent years, a trajectory that has continued even as awareness campaigns have multiplied.
What is striking from a linguistic standpoint is not just how many people are targeted, but how similar the language is across an enormous number of independently reported cases. That consistency is itself a clue. It suggests scripting, not improvisation.
"Love bombing" usually gets discussed as a psychological tactic, overwhelming someone with affection to accelerate trust. But it has a distinct linguistic signature too, one worth learning to recognize on the page, not just in how it feels.
Genuine affection, even when it develops quickly, tends to be specific. It references something the other person actually said, a particular detail, a particular joke, a particular flaw mentioned in passing. Scripted affection tends to be generic and transferable, language that could be copied into a conversation with any target and still function. Terms of endearment appear unusually early and unusually often. Compliments focus on broad, universally flattering qualities, kindness, beauty, intelligence, rather than anything specific to the actual conversation that's happened so far.
A useful test: could this exact sentence have been sent to someone else, with only the name changed, and still make sense? If yes, that's worth noticing.
Every romance scam eventually has to explain why a video call, a phone call, or an in-person meeting isn't possible right now. Over time, a fairly small set of recurring excuses has been documented across thousands of independent cases: deployed military service, an offshore oil rig, a construction contract in a remote location, a medical emergency involving a family member abroad. What's linguistically interesting isn't any single excuse, it's how narrow the actual set of excuses turns out to be once you look across a large number of cases, and how consistently they cluster around occupations that plausibly explain both isolation and unreliable communication.
The language used to deliver these excuses tends to share a shape regardless of the specific cover story: an apology, a plausible-sounding obstacle, and a promise that the obstacle is temporary. The specific details change, deployment orders, oil rig contracts, an emergency abroad, but that three-part structure rarely does.
One of the more subtle patterns is language that gradually positions the scammer as the only person who truly understands the target. This rarely appears as an explicit statement. It shows up as small, repeated framing: gentle skepticism about how much friends or family "really" understand the relationship, or reassurance that it's fine to keep the relationship private "for now." None of this reads as manipulative in isolation. It's the accumulation, over weeks, that does the work.
This matters because it's often the first thing a concerned friend or family member notices from the outside, even before any request for money appears. If someone in a new online relationship starts deflecting questions from people who know them well, that's frequently a downstream effect of language the target has already been absorbing for some time.
Not every romance scam originates from a non-English-speaking operation, but a meaningful share do, and when they do, the same machine-translation fingerprints show up here as in other forms of scam communication. Grammatically correct but slightly unnatural phrasing. Idioms translated too literally. A rhythm to the sentences that doesn't quite match how a native English speaker would actually phrase things, even when every individual word is correct.
This is a genuinely useful thing to know how to spot, not because it proves anything on its own, but because it's one more data point. A profile claiming to be a native English speaker from a specific U.S. city, paired with phrasing that reads like a careful translation, is worth a second look.
None of these patterns are proof on their own. Real people are sometimes awkward, sometimes affectionate quickly, sometimes genuinely unable to video call. The value of knowing these patterns isn't catching a single red flag and declaring fraud. It's noticing when several of them stack up at once, generic affection, a suspiciously convenient reason for no video calls, gentle distancing from people who'd ask hard questions, phrasing that feels just slightly translated. Any one of those alone is unremarkable. All of them together, in the same relationship, is a pattern worth taking seriously.
The language, once you know what to listen for, is usually the first thing that gives it away, well before any money is ever mentioned.