By
Stuart Kerr, Technology Correspondent, LiveAIWire
Romance fraud has always exploited human vulnerability. What has
changed is the scale at which it can now be executed and the sophistication
of the tools available to perpetrators. Artificial intelligence has
transformed online dating scams from a cottage industry requiring significant
manual effort into an industrialised operation capable of maintaining
thousands of fake relationships simultaneously, adapting in real time to the
responses of targets, and generating convincing audio, video, and imagery
that no longer requires specialist skills to produce.
The FBI’s Internet Crime Complaint Centre reported losses
exceeding 1.3 billion dollars to romance scams in the United States in 2022.
In the UK, Action Fraud recorded over 8,000 romance fraud reports in 2023
with losses exceeding 92 million pounds. These are not victimless crimes: the
financial losses are often devastating, and the psychological damage to those
who believed they were in a genuine relationship can be severe and
lasting.
The Generative AI Upgrade to Romance Fraud
Traditional romance scams required human operators to maintain
consistent personas across extended periods, construct plausible backstories,
and manage the emotional dynamics of multiple targets simultaneously.
Generative AI has removed most of those constraints. Large language models
can sustain contextually coherent conversations indefinitely, remember
details shared by the target, and adjust emotional register in response to
cues in the target’s messages. What once required a skilled human manipulator
can now be automated at scale, allowing operations that once ran perhaps a
hundred concurrent fraudulent relationships to manage
thousands.
The visual dimension has evolved equally dramatically. AI image
generation tools create photorealistic profile photographs of people who do
not exist, tailored to whatever demographic the target is likely to find
attractive. The Europol
report on AI-enabled crime published in 2024 identified deepfake
video in romance fraud as an emerging and rapidly growing threat vector,
noting that video calls, once considered a reliable verification method, can
now be faked with consumer-grade hardware.
Faith Communities: A Specific Vulnerability
Religious communities often emphasise trust, charity, and the
importance of helping others in need. The values that make faith communities admirable
also create vulnerabilities that fraudsters have learned to exploit
systematically. Scam operations targeting religious dating apps and
faith-based social networks have been documented across multiple
denominations in North America, the UK, and Australia. Fraudsters adopt
religious personas, use faith-appropriate language and scriptural references,
and build relationships over weeks or months before the financial request
materialises, typically framed in religious terms: a mission trip, a medical
emergency, a business venture they want to share with someone they
trust.
What This Means for You
Protecting yourself against AI-enhanced romance fraud requires
updating assumptions formed in an era of less sophisticated scams. The tests
that once provided reasonable assurance, asking for a video call, checking
profile photographs in reverse-image searches, looking for conversation
inconsistencies, are all less reliable than they were. The most reliable
protective heuristic is one that predates AI: extraordinary financial
requests from people you have never met in person should be declined
regardless of how well-established the online relationship feels. The
strength of an emotional connection formed entirely through digital
communication is not evidence of the communicant’s
authenticity.
As LiveAIWire has covered in analysis of AI
and dark network operations, criminal infrastructure supporting
romance fraud increasingly overlaps with broader organised crime networks,
and the proceeds fund other criminal activities. The operations themselves
have become industrial enterprises rather than individual opportunistic
crimes.
Platform Responsibility and AI Detection
Dating platforms have significant responsibilities in this space
and are increasingly deploying AI to meet them. Profile verification systems
using computer vision detect AI-generated photographs. Behavioural analysis
systems flag accounts that send identical messages to large numbers of users.
Natural language processing detects financial solicitation language before
money changes hands. The effectiveness varies: large, well-resourced
platforms have invested substantially in trust and safety AI, while smaller
platforms and those serving niche communities often lack comparable
capability.
The Federal
Trade Commission’s research on romance scam tactics provides
detailed documentation of the patterns most associated with fraud, feeding
into training datasets used by platform detection systems. The adversarial
dynamic is real: as platforms improve detection, fraud operations adapt their
tactics to avoid triggering the indicators that trained detection models look
for.
The Psychology of AI-Mediated Manipulation
Research in human-computer interaction has documented cases in
which users of AI companion applications developed genuine emotional
attachment to systems they knew to be artificial. In the context of romance
fraud, where targets believe they are interacting with a real person, the
emotional engagement is deeper and the psychological consequences of
discovering the deception more severe. Mental health support organisations
including the National Cyber Security Centre in the UK have developed
specific resources for romance fraud victims, recognising that grief and
shame following discovery can be comparable to that following a genuine
relationship breakdown.
The NCSC
guidance on avoiding romance fraud is an accessible starting point
for anyone concerned about the authenticity of an online relationship. The UK
Online Safety Act imposes duties on platforms to protect users from fraud,
and Ofcom is developing codes of practice specifying required protective
measures. As LiveAIWire has reported in coverage of AI
fraud in financial services, the cross-border dimension of
AI-enabled fraud is one of the central challenges facing enforcement
agencies: criminal networks are based in jurisdictions with limited
extradition cooperation, while victims span the globe.
What Regulation Can and Cannot Achieve
Legislative and regulatory responses to AI-enhanced romance fraud
are developing, though lagging behind the technology. The UK Online Safety
Act imposes duties on platforms to protect users from fraud, including
romance fraud, and Ofcom is developing detailed codes of practice specifying
required protective measures with fines of up to 10 percent of global annual
turnover for non-compliance. In the United States, the FTC has increased
enforcement action against platforms that fail to implement adequate fraud
prevention measures.
The fundamental challenge for regulators is the global nature of
romance fraud operations. Criminal networks running these schemes are
typically based in jurisdictions with limited extradition cooperation and
weak enforcement of cybercrime laws. The platforms that host the
conversations may be registered in one country, operated from another, and
serving victims in a third. Regulation of platforms within reachable
jurisdictions is valuable, but it addresses only part of an ecosystem whose
most harmful elements are specifically structured to evade regulatory
reach.
Technology companies whose platforms are exploited for romance
fraud face a genuine tension between user experience and fraud prevention.
Verification measures that effectively prevent fraudulent account creation
also create friction for genuine users. Platforms competing for user growth
have commercial incentives to minimise that friction, which is one reason
that fraud prevention investments are often insufficient absent regulatory
pressure. The Online Safety Act’s liability shift, which imposes meaningful
financial consequences on platforms for fraud-related harms that they failed
to take reasonable steps to prevent, is designed to change this incentive
structure. Whether it succeeds depends substantially on Ofcom’s willingness
to pursue enforcement action against well-resourced
platforms.
Protecting Yourself: Practical Steps
Beyond awareness of how the scams operate, there are practical
verification steps worth internalising. Any person you meet online who claims
a profession that explains why they cannot meet in person, such as military
service overseas, offshore oil work, or international medical missions, is
using a script that romance fraud operators have refined over years. These
backstories are chosen precisely because they provide a plausible explanation
for why a video call might be difficult and why a financial emergency might
be impossible to handle through local channels.
Speaking with someone who knows you well about a new online
relationship is one of the most effective protective measures available.
Fraud victims frequently report that the scammer actively discouraged them
from discussing the relationship with friends or family, framing the secrecy
as romantic exclusivity. This isolation is a deliberate tactic: people who
discuss their online relationships with trusted others are far more likely to
receive sceptical feedback that prompts reflection before financial loss
occurs. If an online partner discourages you from talking about them, that is
a warning sign in itself.
About the Author
Stuart Kerr is the Technology Correspondent at LiveAIWire,
covering artificial intelligence across society, policy, and industry. About
LiveAIWire.