Romance fraud has always relied on a constructed identity. What has changed is how convincingly that identity can now be built and defended in real time, and what that means for the advice we give people trying to verify who they are really speaking to.
Intelligence briefAI & Synthetic IntelligencePersonal Intelligence
For years, the standard defence against romance fraud was simple: ask for a video call. A scammer working from a folder of stolen photographs could not produce a live, responsive video of a person who did not exist. That single test caught a meaningful share of fraudulent approaches.
That test is weakening. The tools available to fabricate a convincing digital identity, in photographs, in voice, and increasingly in real-time video, have become more accessible, and the old advice needs updating rather than repeating.
Three developments matter more than the others.
Synthetic profile photographs. AI-generated faces are no longer identifiable by the tell-tale flaws that made earlier synthetic images easy to spot, symmetric earrings, garbled text in the background, an unnatural blur at the hairline. Detection now requires structured technical analysis rather than a visual once-over, which is precisely why photo provenance checks belong to an investigation, not a casual glance.
Voice cloning and scripted long cons. Short voice clips are now sufficient to generate a passable synthetic voice, and combined with a scripted, chatbot-assisted conversation style, this has made the “too good to be true, too consistent to be human” tell less reliable than it used to be.
Live video workarounds. The video-call test has not disappeared, but it is no longer unassailable. Reports across multiple jurisdictions have documented cases where video calls were manipulated in real time or where the request for a call was deflected using a constructed reason that itself fits a known pattern, a supposed camera fault, a conveniently timed connectivity issue, a claimed deployment or location that makes video verification “impossible.”
What has stayed constant is the underlying behavioural pattern that romance fraud has always followed, and this is still the more reliable signal:
None of these signals require any technical sophistication to recognise. They are why a structured assessment looks at the whole pattern of behaviour and communication, not just the authenticity of a photograph.
A growing share of romance fraud now converges with investment fraud. The relationship itself becomes the delivery mechanism for an introduction to a cryptocurrency “opportunity,” often presented as something the other person is doing well from and wants to share, rather than an outright request for money. This framing is more persuasive than a direct ask, and it means a verification exercise increasingly needs to look at claimed investment activity alongside the relationship itself.
A structured verification does not stop at the photograph. It tests the specific claims a person has made, ordinarily against documented public sources, and maps the relationship’s behavioural pattern against the typologies above. Where photographs or video are involved, findings on authenticity are worded as consistent with, not proof of, a given pattern, because detection tooling in this space is probabilistic, and overstating certainty does not survive scrutiny.
This is the basis of our AI Romance Fraud Verification service, sitting within our AI & Synthetic Intelligence capability.
If a relationship that began online is moving quickly, resists straightforward verification, and involves a financial or investment element, the tools to check it properly now need to go further than they used to. The behavioural pattern is still the most reliable signal available, but the digital evidence around it can no longer be taken at face value without structured verification.