Scammers who use dating sites to trick people into handing over cash can be spotted using artificial intelligence, research suggests.
A neural network has analysed profiles, messages and images from real dating data to get better at spotting fakes.
It sampled age, gender and ethnicity as well as the language people use to describe themselves.
The system proved accurate at spotting scammers and fakes in 93% of cases, the researchers said.
Statistics gathered by the UK's police reporting centre Action Fraud suggest British people lost £50m to romance scams in 2018. About 63% of the victims were women.
Computer scientists in the UK, US and Australia collaborated on the AI-based system, which found that those making fake profiles were more likely to be men (60%) and had an average age of 50.
Profiles of fakes and scammers used more images and "emotive language", they said. The common words employed were "caring", "passionate" and "loving".
The system was trained using almost 15,000 profiles from the free Dating 'N More website. The computer science project used data from the service because it publicly posts fake profiles when they are discovered.
Ultimately the team hopes to create an early warning system that can spot scammers as they set up accounts and begin the process of contacting victims.
The researchers said scams on dating sites and apps were hard to tackle because they were usually not large campaigns and were not generated automatically.
The researchers suggested their method could be harder to get around than some current approaches, which rely on blacklists and other basic technical tricks to thwart repeat offenders.
They added: "We aim to more broadly examine the available data on online dating fraud, seeking information actionable for enforcement and other countermeasures."